Research Publications
Browse peer-reviewed journal articles, conference papers, and scholarly publications. Use the filters below to refine results by type, author, year, or research area.
1.
Faisal, Syed; Muniyandy, Elangovan; Lokesh, S.; Jayanthi, M.
In: 2025, ISBN: 2731-4820.
@article{606,
title = {A Random Graph Diffusion Attention Network with Great Wall Construction and Clinical Metadata for Improved Kidney Cancer Diagnosis and Surgical Planning},
author = {Syed Faisal and Elangovan Muniyandy and S. Lokesh and M. Jayanthi},
url = {https://link.springer.com/article/10.1007/s44174-025-00447-6},
doi = {10.1007/s44174-025-00447-6},
isbn = {2731-4820},
year = {2025},
date = {2025-08-01},
abstract = {Kidney cancer develops through abnormal cell multiplication within the renal cortex or pelvis and often results from smoking combined with obesity and hypertension as well as genetic predisposition. The condition produces three main signs: hematuria, flank pain and weight loss. Current deep learning detection algorithms show low accuracy and high error rates. The Random Graph Diffusion Attention Network with Great Wall Construction (RGDAN-GWCA) method aims to enhance detection accuracy and optimize classification performance for solving this issue. This study uses RGDAN-GWCA to detect and Surgical Planning the kidney cancer. This proposed approach integrates CT images with clinical data for its operation. The proposed method adopts KiTS21 because it represents a standalone dataset. The proposed method solves major diagnostic obstacles for kidney cancer through advanced image preprocessing methods and extraction techniques and classification algorithms. The analysis of CT images occurred with Gradient Domain Guided Filtering (GDGF) and the Spike-driven transformer method implemented metadata refinements. The method Inverse Z-transform and Wiener Hopf Factorization (InZ-Tr-WHF) for CT image feature extraction with Min Max Normalization processing of clinical metadata. The RGDAN-GWCA system used a Great Wall Construction Algorithm to combine and analyze extracted features for performing accurate patient classification. This proposed methodology delivers outstanding results for all performance measurement criteria which include 99.65% accuracy along with 99.61% precision, 99.63% recall, 99.56% specificity and 99.62% F1-score. Clinical assessment of tumor volume alongside cancer stage established themselves as the key medical indicators used by surgeons for making treatment decisions. The surgical planning method shows remarkable promise to assist doctors in deciding appropriate nephrectomy treatments for patients dealing with kidney cancer.},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Kidney cancer develops through abnormal cell multiplication within the renal cortex or pelvis and often results from smoking combined with obesity and hypertension as well as genetic predisposition. The condition produces three main signs: hematuria, flank pain and weight loss. Current deep learning detection algorithms show low accuracy and high error rates. The Random Graph Diffusion Attention Network with Great Wall Construction (RGDAN-GWCA) method aims to enhance detection accuracy and optimize classification performance for solving this issue. This study uses RGDAN-GWCA to detect and Surgical Planning the kidney cancer. This proposed approach integrates CT images with clinical data for its operation. The proposed method adopts KiTS21 because it represents a standalone dataset. The proposed method solves major diagnostic obstacles for kidney cancer through advanced image preprocessing methods and extraction techniques and classification algorithms. The analysis of CT images occurred with Gradient Domain Guided Filtering (GDGF) and the Spike-driven transformer method implemented metadata refinements. The method Inverse Z-transform and Wiener Hopf Factorization (InZ-Tr-WHF) for CT image feature extraction with Min Max Normalization processing of clinical metadata. The RGDAN-GWCA system used a Great Wall Construction Algorithm to combine and analyze extracted features for performing accurate patient classification. This proposed methodology delivers outstanding results for all performance measurement criteria which include 99.65% accuracy along with 99.61% precision, 99.63% recall, 99.56% specificity and 99.62% F1-score. Clinical assessment of tumor volume alongside cancer stage established themselves as the key medical indicators used by surgeons for making treatment decisions. The surgical planning method shows remarkable promise to assist doctors in deciding appropriate nephrectomy treatments for patients dealing with kidney cancer.
2.
Madhavan, Sowmya; Reddy, Satti; Kavitha, T.; kaleem, Afshan; Shilpa, V.
Energy-Efficient Reliable Data Transmission Using Optimized Cyclone Foraging Strategy in 5G Wireless Sensor Networks Journal Article
In: Internet Technology Letters, vol. 8, pp. e643+, 2025, ISBN: 2476-1508.
@article{394,
title = {Energy-Efficient Reliable Data Transmission Using Optimized Cyclone Foraging Strategy in 5G Wireless Sensor Networks},
author = {Sowmya Madhavan and Satti Reddy and T. Kavitha and Afshan kaleem and V. Shilpa},
url = {https://doi.org/10.1002/itl2.643},
doi = {10.1002/itl2.643},
isbn = {2476-1508},
year = {2025},
date = {2025-03-01},
journal = {Internet Technology Letters},
volume = {8},
pages = {e643+},
publisher = {John Wiley & Sons, Ltd},
abstract = {As a fundamental supporting technology of 5th Generation (5G) systems, wireless sensor networks (WSN) are handling a new challenge to enhance its energy-efficient reliable transmission. However, energy usage and network lifetime are observed as challenging tasks because of limited battery capacity and open environments. The cyclone foraging strategy with beluga whale optimization (CFS-BWO) is proposed for energy-efficient reliable data transmission for 5G WSN. The CFS is applied for improving the exploitation phase of traditional BWO, where population transfers in spiral orientation among the best solutions. CFS-BWO optimizes the routing and data aggregation process, which minimizes the energy usage and enhances the network lifetime. The average delay, residual energy, communication cost, and distance are adopted as fitness functions for optimizing the best solution in both CH and route path selection. The performance was calculated by the metrics of residual energy, packet delivery ratio (PDR), and delay across 200, 400, 600, 800, and 1000 rounds. The CFS-BWO reaches residual energy of 0.87?J, PDR of 0.98, and delay of 15?ms for 1000 rounds when compared to optimal cluster-based routing (Optimal-CBR).},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
As a fundamental supporting technology of 5th Generation (5G) systems, wireless sensor networks (WSN) are handling a new challenge to enhance its energy-efficient reliable transmission. However, energy usage and network lifetime are observed as challenging tasks because of limited battery capacity and open environments. The cyclone foraging strategy with beluga whale optimization (CFS-BWO) is proposed for energy-efficient reliable data transmission for 5G WSN. The CFS is applied for improving the exploitation phase of traditional BWO, where population transfers in spiral orientation among the best solutions. CFS-BWO optimizes the routing and data aggregation process, which minimizes the energy usage and enhances the network lifetime. The average delay, residual energy, communication cost, and distance are adopted as fitness functions for optimizing the best solution in both CH and route path selection. The performance was calculated by the metrics of residual energy, packet delivery ratio (PDR), and delay across 200, 400, 600, 800, and 1000 rounds. The CFS-BWO reaches residual energy of 0.87?J, PDR of 0.98, and delay of 15?ms for 1000 rounds when compared to optimal cluster-based routing (Optimal-CBR).
3.
Benchmarking of Machine Learning for Anomalybased Intrusion Detection Systems Using LSTM-RNN Miscellaneous
2025, ISBN: 9798331536770 (ISBN), (0).
@misc{729,
title = {Benchmarking of Machine Learning for Anomalybased Intrusion Detection Systems Using LSTM-RNN},
url = {https://ieeexplore.ieee.org/document/11210998},
doi = {10.1109/IACIS65746.2025.11210998},
isbn = {9798331536770 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Over the past few years, the challenge has been the increasing and significant attacks on anomaly detection processes. While attacks in anomaly detection can be easily predicted using Intrusion Detection Systems (IDS), the accuracy of the prediction process remains low. To address the issues associated with Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNN) for IDS detection. Furthermore, Z-score normalization aims to eliminate duplicate data and minimize unknown data during the preprocessing stage. Additionally, the Grasshopper Optimization Algorithm is used to select relevant features from the margin. Behavior analysis is employed to verify each type of data in a prediction dataset and identify the necessary checks at each performance stage. Finally, the proposed method evaluates testing and training values, classifies intrusions, and detects various attacks in the early stages. The proposed technique reduces time complexity and improves the accuracy to 93%.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {misc}
}
Over the past few years, the challenge has been the increasing and significant attacks on anomaly detection processes. While attacks in anomaly detection can be easily predicted using Intrusion Detection Systems (IDS), the accuracy of the prediction process remains low. To address the issues associated with Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNN) for IDS detection. Furthermore, Z-score normalization aims to eliminate duplicate data and minimize unknown data during the preprocessing stage. Additionally, the Grasshopper Optimization Algorithm is used to select relevant features from the margin. Behavior analysis is employed to verify each type of data in a prediction dataset and identify the necessary checks at each performance stage. Finally, the proposed method evaluates testing and training values, classifies intrusions, and detects various attacks in the early stages. The proposed technique reduces time complexity and improves the accuracy to 93%.
4.
Shirisha, M. S.; Prasad, M. S.; Naikodi, C.; Srinivas, B. G.; Rao, Bhaskara
Secured IoT Data Management Using AES Encryption And Blockchain Technology Journal Article
In: Journal of Information Systems Engineering and Management, vol. 10, pp. 728-735,, 2025, ISBN: 24684376 (ISSN), (0).
@article{378,
title = {Secured IoT Data Management Using AES Encryption And Blockchain Technology},
author = {M. S. Shirisha and M. S. Prasad and C. Naikodi and B. G. Srinivas and Bhaskara Rao},
doi = {10.52783/jisem.v10i9s.1299},
isbn = {24684376 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Journal of Information Systems Engineering and Management},
volume = {10},
pages = {728-735,},
publisher = {IADITI - International Association for Digital Transformation and Technological Innovation},
abstract = {The swift expansion of IoT devices necessitates robust mechanisms to protect sensitive data from unauthorized access and tampering. This paper fulfills this requirement by integrating AES encryption and blockchain technology. IoT data is encrypted using dynamically generated private keys with AES in CBC mode, ensuring confidentiality, while a Flask application facilitates secure data processing. The encrypted data and metadata are stored on a Solidity-based smart contract, providing decentralized and tamper-proof storage. Results demonstrate enhanced data security, immutability, and traceability, making the framework scalable and suitable for modern IoT ecosystems. Analysis confirms the effectiveness of combining encryption and blockchain to mitigate security risks.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
The swift expansion of IoT devices necessitates robust mechanisms to protect sensitive data from unauthorized access and tampering. This paper fulfills this requirement by integrating AES encryption and blockchain technology. IoT data is encrypted using dynamically generated private keys with AES in CBC mode, ensuring confidentiality, while a Flask application facilitates secure data processing. The encrypted data and metadata are stored on a Solidity-based smart contract, providing decentralized and tamper-proof storage. Results demonstrate enhanced data security, immutability, and traceability, making the framework scalable and suitable for modern IoT ecosystems. Analysis confirms the effectiveness of combining encryption and blockchain to mitigate security risks.
5.
Prasad, S. M. S.; Shirisha, G. M. S.; Naikodi, C.; Srinivas, B. G.; Rao, B. B.
A Comparative Study of Sequence of Multiple Fingerprints for Secured Authentication Journal Article
In: Journal of Information Systems Engineering and Management, vol. 10, pp. 698-708,, 2025, ISBN: 24684376 (ISSN), (0).
@article{381,
title = {A Comparative Study of Sequence of Multiple Fingerprints for Secured Authentication},
author = {S. M. S. Prasad and G. M. S. Shirisha and C. Naikodi and B. G. Srinivas and B. B. Rao},
doi = {10.52783/jisem.v10i9s.1296},
isbn = {24684376 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Journal of Information Systems Engineering and Management},
volume = {10},
pages = {698-708,},
publisher = {IADITI - International Association for Digital Transformation and Technological Innovation},
abstract = {Smartphones are becoming increasingly prevalent and most of them use fingerprint recognition to authenticate any application, from financial transactions to login. Imposters are attacking single fingerprint template quite easily. So, next security level for smartphones needs to be implemented in order to strengthen existing method. One such method is implemented by employing Sequence of Multiple Fingerprints (SMF). This paper presents a lightweight, cost-effective application based sequential fingerprint authentication technique when compared with other modern techniques. The proposed system is designed for environments with limited computational resources, offering enhanced security and efficiency. Unlike traditional single-fingerprint authentication methods, our algorithm employs sequential fingerprint input for improved accuracy and robustness. Experimental results demonstrate a low False Acceptance Rate (FAR) of 0.5%–3% and a False Rejection Rate (FRR) of 1.8%–5%, with significantly reduced execution costs and processing times compared to existing methods. The system is ideal for Smartphones, IoT applications, including access control and smart lock systems, where lightweight and scalable solutions are essential.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Smartphones are becoming increasingly prevalent and most of them use fingerprint recognition to authenticate any application, from financial transactions to login. Imposters are attacking single fingerprint template quite easily. So, next security level for smartphones needs to be implemented in order to strengthen existing method. One such method is implemented by employing Sequence of Multiple Fingerprints (SMF). This paper presents a lightweight, cost-effective application based sequential fingerprint authentication technique when compared with other modern techniques. The proposed system is designed for environments with limited computational resources, offering enhanced security and efficiency. Unlike traditional single-fingerprint authentication methods, our algorithm employs sequential fingerprint input for improved accuracy and robustness. Experimental results demonstrate a low False Acceptance Rate (FAR) of 0.5%–3% and a False Rejection Rate (FRR) of 1.8%–5%, with significantly reduced execution costs and processing times compared to existing methods. The system is ideal for Smartphones, IoT applications, including access control and smart lock systems, where lightweight and scalable solutions are essential.
6.
Devi, K. K.; Kumar, J. P.; Karanth, S.
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 979-835035623-6 (ISBN), (0).
@proceedings{417,
title = {Optimized Method for Prediction and Recommendation of Crops Using Fusion Ensemble Learning Crop Recommendation Model},
author = {K. K. Devi and J. P. Kumar and S. Karanth},
url = {https://ieeexplore.ieee.org/document/10958850},
doi = {10.1109/ICAECT63952.2025.10958850},
isbn = {979-835035623-6 (ISBN)},
year = {2025},
date = {2025-01-01},
journal = {2025 5th International Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies, ICAECT 2025},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Fusion Ensemble Learning Crop Recommendation Model introduces an advanced crop recommendation system utilizing machine learning (ML), Robotic Process Automation (RPA) with UiPath, and Explainable AI (XAI). The system aggregates historical and real-time IoT data on weather and soil nutrient content to provide optimized recommendations for crops and NPK (Nitrogen, Phosphorus, and Potassium). By creating an FELCR (Fusion Ensemble Learning Crop Recommendation model) using Random Forest, CatBoost, and XGBoost models, the system achieves robust and accurate crop predictions under various environmental conditions. UiPath automates the collection of this data, streamlining the process and ensuring farmers have timely inputs. To enhance transparency, the model incorporates XAI techniques, specifically LIME (Local Interpretable Model-agnostic Explanations), which allows farmers to understand and trust the recommendations, making the system highly user friendly. The ensemble model demonstrated an accuracy of 99.70%, and its recommendations have been widely adopted by farmers. This study aligns with the United Nations Sustainable Development Goals (SDGs), particularly Goal 2, promoting sustainable and efficient agricultural practices.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Fusion Ensemble Learning Crop Recommendation Model introduces an advanced crop recommendation system utilizing machine learning (ML), Robotic Process Automation (RPA) with UiPath, and Explainable AI (XAI). The system aggregates historical and real-time IoT data on weather and soil nutrient content to provide optimized recommendations for crops and NPK (Nitrogen, Phosphorus, and Potassium). By creating an FELCR (Fusion Ensemble Learning Crop Recommendation model) using Random Forest, CatBoost, and XGBoost models, the system achieves robust and accurate crop predictions under various environmental conditions. UiPath automates the collection of this data, streamlining the process and ensuring farmers have timely inputs. To enhance transparency, the model incorporates XAI techniques, specifically LIME (Local Interpretable Model-agnostic Explanations), which allows farmers to understand and trust the recommendations, making the system highly user friendly. The ensemble model demonstrated an accuracy of 99.70%, and its recommendations have been widely adopted by farmers. This study aligns with the United Nations Sustainable Development Goals (SDGs), particularly Goal 2, promoting sustainable and efficient agricultural practices.
7.
Khan, I.; Somshekhar, D.; Harikrishnan, N.; Neetha, N.; Pavithra, D.
Hierarchical Meta-Reinforcement Learning for Uncertainty-Aware Resource Allocation in C-V2X Networks Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 979-833150574-5 (ISBN), (0).
@proceedings{432,
title = {Hierarchical Meta-Reinforcement Learning for Uncertainty-Aware Resource Allocation in C-V2X Networks},
author = {I. Khan and D. Somshekhar and N. Harikrishnan and N. Neetha and D. Pavithra},
url = {https://ieeexplore.ieee.org/document/10968190},
doi = {10.1109/ICMLAS64557.2025.10968190},
isbn = {979-833150574-5 (ISBN)},
year = {2025},
date = {2025-01-01},
journal = {2nd International Conference on Machine Learning and Autonomous Systems, ICMLAS 2025 - Proceedings},
pages = {167-172,},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Ensuring smooth data interchange between vehicles and infrastructure depends on the smart use of communication resources, including bandwidth, power, and time slots. The framework must respond to user specific requirements, adjust to changing network conditions, and maximize efficiency while minimizing interference. Nevertheless, it has several challenges, such as limited resources, excessive energy consumption, and delays caused due to frequent topological changes by vehicle mobility. This paper proposes the Hierarchical Meta Reinforcement Learning framework based on Information Volume Evidential Markov Decision Processes to address these challenges. Unlike conventional reinforcement learning techniques, this methodology uses hierarchical reinforcement learning to make adaptive decisions in challenging situations. Additionally, by utilizing previously learned policies, it uses meta-learning to facilitate quick tasks adaption, increasing operational efficiency and flexibility. The proposed method achieves a highest V2V link success probability of 98% for 100 vehicles outperforming other techniques like MA-DDQN.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Ensuring smooth data interchange between vehicles and infrastructure depends on the smart use of communication resources, including bandwidth, power, and time slots. The framework must respond to user specific requirements, adjust to changing network conditions, and maximize efficiency while minimizing interference. Nevertheless, it has several challenges, such as limited resources, excessive energy consumption, and delays caused due to frequent topological changes by vehicle mobility. This paper proposes the Hierarchical Meta Reinforcement Learning framework based on Information Volume Evidential Markov Decision Processes to address these challenges. Unlike conventional reinforcement learning techniques, this methodology uses hierarchical reinforcement learning to make adaptive decisions in challenging situations. Additionally, by utilizing previously learned policies, it uses meta-learning to facilitate quick tasks adaption, increasing operational efficiency and flexibility. The proposed method achieves a highest V2V link success probability of 98% for 100 vehicles outperforming other techniques like MA-DDQN.
8.
Hoovayya, Uluvaru; Kumar, J. P.
Improved LinkNet-DenseNet architecture with fine-tuned parameters for abnormal human activity detection for video surveillance Journal Article
In: Signal, Image and Video Processing, vol. 19, 2025, ISBN: 18631703 (ISSN), (0).
@article{445,
title = {Improved LinkNet-DenseNet architecture with fine-tuned parameters for abnormal human activity detection for video surveillance},
author = {Uluvaru Hoovayya and J. P. Kumar},
url = {https://link.springer.com/article/10.1007/s11760-025-04343-w},
doi = {10.1007/s11760-025-04343-w},
isbn = {18631703 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Signal, Image and Video Processing},
volume = {19},
publisher = {Springer Science and Business Media Deutschland GmbH},
abstract = {Video-based human abnormal activity detection is to identify and understand abnormal behaviors and actions from the video sequence. For this purpose, an Improved LinkNet-DenseNet (ILN-DN) architecture-based abnormal human activity detection model is introduced in this work, which includes 4 working stages. The novelty of the work lies in the integration of advanced segmentation techniques, enhanced feature extraction methods, and a hybrid classification model for abnormal human activity detection. Additionally, the introduction of a modified optimization algorithm significantly improves convergence speed and detection accuracy. In the pre-processing stage, median filtering is applied to get the pre-processed image (frame), after converting the input video into frames. Afterwards, that pre-processed image gets segmented by the Improved Mask R-CNN model. From the segmented image, features such as Mobile-Based Scale Invariant Feature Transform (MoBSIFT), Improved Shape Local Binary Texture (SLBT), and Hierarchy of Skeleton are extracted. Finally, based on the extracted features, abnormal human activity is detected effectively, with the utilization of a proposed hybrid classification model namely Improved LinkNet-DenseNet (ILN-DN) architecture. The DenseNet architecture is chosen for its efficient feature reuse through dense connections, which enhances the model’s ability to learn complex patterns and improve performance in detecting subtle anomalies. LinkNet, known for its strong segmentation capabilities, is utilized to precisely isolate regions of interest, such as human figures, from the background. Combining both architectures in the ILN-DN model leverages DenseNet’s deep feature extraction and LinkNet’s segmentation precision, making it highly effective for abnormal human activity detection in video surveillance. The parameters of these classifiers are fine-tuned with the utilization of the Improved Red Panda Optimization (IRPO) algorithm, to enhance the detection performance.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Video-based human abnormal activity detection is to identify and understand abnormal behaviors and actions from the video sequence. For this purpose, an Improved LinkNet-DenseNet (ILN-DN) architecture-based abnormal human activity detection model is introduced in this work, which includes 4 working stages. The novelty of the work lies in the integration of advanced segmentation techniques, enhanced feature extraction methods, and a hybrid classification model for abnormal human activity detection. Additionally, the introduction of a modified optimization algorithm significantly improves convergence speed and detection accuracy. In the pre-processing stage, median filtering is applied to get the pre-processed image (frame), after converting the input video into frames. Afterwards, that pre-processed image gets segmented by the Improved Mask R-CNN model. From the segmented image, features such as Mobile-Based Scale Invariant Feature Transform (MoBSIFT), Improved Shape Local Binary Texture (SLBT), and Hierarchy of Skeleton are extracted. Finally, based on the extracted features, abnormal human activity is detected effectively, with the utilization of a proposed hybrid classification model namely Improved LinkNet-DenseNet (ILN-DN) architecture. The DenseNet architecture is chosen for its efficient feature reuse through dense connections, which enhances the model’s ability to learn complex patterns and improve performance in detecting subtle anomalies. LinkNet, known for its strong segmentation capabilities, is utilized to precisely isolate regions of interest, such as human figures, from the background. Combining both architectures in the ILN-DN model leverages DenseNet’s deep feature extraction and LinkNet’s segmentation precision, making it highly effective for abnormal human activity detection in video surveillance. The parameters of these classifiers are fine-tuned with the utilization of the Improved Red Panda Optimization (IRPO) algorithm, to enhance the detection performance.
9.
Jayasimha, S. R.; Patil, R. V.; Yashashwini, S.; Prabu, M.; Srisathirapathy, S.; Maranan, R.
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 979-833151224-8 (ISBN), (0).
@proceedings{447,
title = {Spilled Deep Capsule Neural Network with Skill Optimization Algorithm for Breast Cancer Recognition in Mammograms},
author = {S. R. Jayasimha and R. V. Patil and S. Yashashwini and M. Prabu and S. Srisathirapathy and R. Maranan},
url = {https://ieeexplore.ieee.org/document/11004784},
doi = {10.1109/ICICT64420.2025.11004784},
isbn = {979-833151224-8 (ISBN)},
year = {2025},
date = {2025-01-01},
journal = {Proceedings of 8th International Conference on Inventive Computation Technologies, ICICT 2025},
pages = {632-637,},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Breast cancer stands as the main reason for cancer deaths among women worldwide so early detection plays an essential role in raising both survival statistics and treatment effectiveness. The common breast cancer screening method known as mammography leads radiologists to struggle when evaluating mammograms because they often make incorrect diagnoses that result in delayed medical procedures. Standard assessment techniques experience difficulties in identifying faint abnormalities which causes both incorrect positive and negative results. The proposed research implements a new methodology that utilizes MIAS dataset to detect breast cancer. The Adaptive Morphological Wavelet Perona-Malik Filter Algorithm operates on mammogram images for preprocessing to optimize quality and save vital image characteristics. A Spilled Deep Capsule Neural Network function (SDCN) utilizes the network for effective mammographic image detection and feature extraction. The Skill Optimization Algorithm (SOA) serves as a parameter optimization tool which boosts both accuracy and efficiency of the model. The developed method succeeded in reaching 99.9% accuracy while surpassing previous standard examination methods. The research demonstrates how advanced deep learning and optimization algorithms support radiologists by detecting breast cancer properly and expeditiously which creates better patient healthcare outcomes.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Breast cancer stands as the main reason for cancer deaths among women worldwide so early detection plays an essential role in raising both survival statistics and treatment effectiveness. The common breast cancer screening method known as mammography leads radiologists to struggle when evaluating mammograms because they often make incorrect diagnoses that result in delayed medical procedures. Standard assessment techniques experience difficulties in identifying faint abnormalities which causes both incorrect positive and negative results. The proposed research implements a new methodology that utilizes MIAS dataset to detect breast cancer. The Adaptive Morphological Wavelet Perona-Malik Filter Algorithm operates on mammogram images for preprocessing to optimize quality and save vital image characteristics. A Spilled Deep Capsule Neural Network function (SDCN) utilizes the network for effective mammographic image detection and feature extraction. The Skill Optimization Algorithm (SOA) serves as a parameter optimization tool which boosts both accuracy and efficiency of the model. The developed method succeeded in reaching 99.9% accuracy while surpassing previous standard examination methods. The research demonstrates how advanced deep learning and optimization algorithms support radiologists by detecting breast cancer properly and expeditiously which creates better patient healthcare outcomes.
10.
Janartanan, V. D.; Venkatesh, R. T.; Rao, P. B. S.; Vaithilingam, S. D.; Srinivasan, A.
Stress and Depression Classification in Social Media using Contextual Knowledge Attention based Gated Recurrent Network Journal Article
In: International Journal of Intelligent Engineering and Systems, vol. 18, pp. 420-431,, 2025, ISBN: 2185310X (ISSN), (0).
@article{457,
title = {Stress and Depression Classification in Social Media using Contextual Knowledge Attention based Gated Recurrent Network},
author = {V. D. Janartanan and R. T. Venkatesh and P. B. S. Rao and S. D. Vaithilingam and A. Srinivasan},
url = {https://inass.org/wp-content/uploads/2024/12/2025043029-2.pdf},
doi = {10.22266/ijies2025.0430.29},
isbn = {2185310X (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {International Journal of Intelligent Engineering and Systems},
volume = {18},
pages = {420-431,},
publisher = {Intelligent Network and Systems Society},
abstract = {The detection of stress and depression on social media focuses on analyzing social media posts to identify signs of stress and tendencies toward depression. It aims to offer early detection of the mental health states. However, detecting signs of stress or depression through social media posts presents significant challenges due to the unstructured nature of the text in the posts, and the diverse language styles used by individuals. So, this research proposes Contextual Knowledge Attention mechanism-based Gated Recurrent Unit (CKA-GRU) for the classification of posts from a popular social media blog site. The CKA mechanism dynamically applies weights to various parts of the text, allowing the GRU to prioritise important words based on stress and depression, which enhances classification accuracy. The effectiveness of the proposed CKA-GRU approach is evaluated on two standard datasets: Dreaddit and Depression_Mixed. These datasets consist of user posts analyzed to detect patterns indicative of stress or depression on social media. The relevant features are extracted from the pre-processed data using the Bag of Words (BoW). Finally, the CKA-GRU approach is employed for the classification of posts into binary classes for enhancing the overall performance of the model. The experimental results demonstrate that the proposed CKA-GRU method attains a commendable accuracy of 91.39% and 95.20% on the Dreaddit and Depression_Mixed datasets, respectively. These results prove that the proposed CKA-GRU approach accomplishes superior outcomes than the existing approaches namely, Bow with Logistic Regression (BoW-LR) and Knowledge-aware and Contrastive Network (KC-Net).},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
The detection of stress and depression on social media focuses on analyzing social media posts to identify signs of stress and tendencies toward depression. It aims to offer early detection of the mental health states. However, detecting signs of stress or depression through social media posts presents significant challenges due to the unstructured nature of the text in the posts, and the diverse language styles used by individuals. So, this research proposes Contextual Knowledge Attention mechanism-based Gated Recurrent Unit (CKA-GRU) for the classification of posts from a popular social media blog site. The CKA mechanism dynamically applies weights to various parts of the text, allowing the GRU to prioritise important words based on stress and depression, which enhances classification accuracy. The effectiveness of the proposed CKA-GRU approach is evaluated on two standard datasets: Dreaddit and Depression_Mixed. These datasets consist of user posts analyzed to detect patterns indicative of stress or depression on social media. The relevant features are extracted from the pre-processed data using the Bag of Words (BoW). Finally, the CKA-GRU approach is employed for the classification of posts into binary classes for enhancing the overall performance of the model. The experimental results demonstrate that the proposed CKA-GRU method attains a commendable accuracy of 91.39% and 95.20% on the Dreaddit and Depression_Mixed datasets, respectively. These results prove that the proposed CKA-GRU approach accomplishes superior outcomes than the existing approaches namely, Bow with Logistic Regression (BoW-LR) and Knowledge-aware and Contrastive Network (KC-Net).
11.
Aparna, N.; Preksha, S.; Fareetha, Raheema; Chithra, S. M.
Secure Gene Profile with AES Encryption and Blockchain Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 979-833150724-4 (ISBN), (0).
@proceedings{465,
title = {Secure Gene Profile with AES Encryption and Blockchain},
author = {N. Aparna and S. Preksha and Raheema Fareetha and S. M. Chithra},
url = {https://ieeexplore.ieee.org/document/11041865},
doi = {10.1109/ICAISS61471.2025.11041865},
isbn = {979-833150724-4 (ISBN)},
year = {2025},
date = {2025-01-01},
journal = {Proceedings of 3rd International Conference on Augmented Intelligence and Sustainable Systems, ICAISS 2025},
pages = {1688-1693,},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {The rapid growth of genomic science calls for the safe storage and communication of private gene information. The classic approach using centralized databases exposes gene information to the possibility of being leaked or hacked into, prompting a strong demand for better security methods. In this paper, a novel method in applying Advanced Encryption Standard (AES) encryption integrated with blockchain is developed. Before uploading the gene datasets to the Ganache blockchain, the system secures them using file hashes that are generated with SHA-256's smart contract combination controlled by MataMask. In this way, the environment is more secure and is able to store the encoded gene datasets in a way that is vitiation resistant, secure, and traceable. There is a user interface through which are subsequently authenticated, encrypted by an administrator, and then stored in blockchain. There are advanced analytics and visualization techniques that can be applied to the gene data without violation privacy. The proposed model shows that it is possible to achieve genome data protection by implementing AES encryption in blockchain for effective data management, eliminating the data security gaps associated with precision medicine and personalized healthcare.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
The rapid growth of genomic science calls for the safe storage and communication of private gene information. The classic approach using centralized databases exposes gene information to the possibility of being leaked or hacked into, prompting a strong demand for better security methods. In this paper, a novel method in applying Advanced Encryption Standard (AES) encryption integrated with blockchain is developed. Before uploading the gene datasets to the Ganache blockchain, the system secures them using file hashes that are generated with SHA-256's smart contract combination controlled by MataMask. In this way, the environment is more secure and is able to store the encoded gene datasets in a way that is vitiation resistant, secure, and traceable. There is a user interface through which are subsequently authenticated, encrypted by an administrator, and then stored in blockchain. There are advanced analytics and visualization techniques that can be applied to the gene data without violation privacy. The proposed model shows that it is possible to achieve genome data protection by implementing AES encryption in blockchain for effective data management, eliminating the data security gaps associated with precision medicine and personalized healthcare.
12.
H, D.; Kumar, J. P.
Enhanced Transfer Learning-Based CNN for Abnormal Human Activity Detection in Video Surveillance Using Spatial-Temporal Features Journal Article
In: Cybernetics and Systems, 2025, ISBN: 01969722 (ISSN), (0).
@article{466,
title = {Enhanced Transfer Learning-Based CNN for Abnormal Human Activity Detection in Video Surveillance Using Spatial-Temporal Features},
author = {D. H and J. P. Kumar},
url = {https://www.tandfonline.com/doi/full/10.1080/01969722.2025.2521708},
doi = {10.1080/01969722.2025.2521708},
isbn = {01969722 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Cybernetics and Systems},
publisher = {Taylor and Francis Ltd.},
abstract = {Video surveillance (VS) is essential in today’s environment. Artificial intelligence (AI) including ML, and DL brought too much technological advancement to the surveillance system. The combination of these technologies assists in distinguishing distinct suspicious acts and behaviors from the real-time surveillance video. Human behavior is unpredictable; therefore, it is challenging to determine whether a behavior is suspicious or not. Thereby, we develop a new model for recognizing the abnormal actions of humans in VS. Initially, the input frames are preprocessed via Improved Wiener Filtering (IWF). As the next step, segmentation is done using the Improved SegNet model (ISegNet). Further, spatial and temporal (S&T) features, Improved Motion Estimation (ME), Color Feature, motion boundary SIFT (MoBSIFT) and Local Gradient Increasing Pattern (LGIP) features are extracted. Finally, the detection of abnormal actions of humans takes place via Improved Transfer Learning based CNN (ITL-CNN). The outcomes from ITL-CNN include Abuse, arrest, arson, assault, road accidents, robbery and shooting.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Video surveillance (VS) is essential in today’s environment. Artificial intelligence (AI) including ML, and DL brought too much technological advancement to the surveillance system. The combination of these technologies assists in distinguishing distinct suspicious acts and behaviors from the real-time surveillance video. Human behavior is unpredictable; therefore, it is challenging to determine whether a behavior is suspicious or not. Thereby, we develop a new model for recognizing the abnormal actions of humans in VS. Initially, the input frames are preprocessed via Improved Wiener Filtering (IWF). As the next step, segmentation is done using the Improved SegNet model (ISegNet). Further, spatial and temporal (S&T) features, Improved Motion Estimation (ME), Color Feature, motion boundary SIFT (MoBSIFT) and Local Gradient Increasing Pattern (LGIP) features are extracted. Finally, the detection of abnormal actions of humans takes place via Improved Transfer Learning based CNN (ITL-CNN). The outcomes from ITL-CNN include Abuse, arrest, arson, assault, road accidents, robbery and shooting.
13.
V, Kesavan M; Kumar, Josephine; Ashwin, Nanda
HBRFE: an enhanced recursive feature elimination model for big data classification Journal Article
In: vol. 14, pp. 3061-3074,, 2025.
@article{600,
title = {HBRFE: an enhanced recursive feature elimination model for big data classification},
author = {Kesavan M V and Josephine Kumar and Nanda Ashwin},
doi = {10.11591/eei.v14i4.9595},
year = {2025},
date = {2025-01-01},
volume = {14},
pages = {3061-3074,},
abstract = {The process of classification in big data is a tedious task due to the large number of volumes, veracity, and variety of the data. Classification of big data pave the path to organize the data and improve the classifier performance. This research article proposed a Hadoop framework based recursive feature elimination-based model called HBFRE for extract significant features from the big data by integrating map and reduce frame work. HBFRE extract the significant features by removing the least and irrelevant features from the dataset by using refined recursive feature elimination (RFE) with map and reduce framework. This method takes the mean of each attribute and find the variance in each instance. The proposed model is evaluated and analyzed by the accuracy performance and time complexity. This research utilized various classifier like artificial neural network (ANN), support vector machine (SVM), random forest (RF), k-nearest neighbors (KNN), and AdaBoost to measure the classification performance on the big data. Proposed HBRFE model is compared with different feature selection like RFE, relief, backwards feature elimination, maximum relevance k-nearest neighbors (MR-KNN), and scalable deep ensemble framework big data classification (SDELF-BDC).},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
The process of classification in big data is a tedious task due to the large number of volumes, veracity, and variety of the data. Classification of big data pave the path to organize the data and improve the classifier performance. This research article proposed a Hadoop framework based recursive feature elimination-based model called HBFRE for extract significant features from the big data by integrating map and reduce frame work. HBFRE extract the significant features by removing the least and irrelevant features from the dataset by using refined recursive feature elimination (RFE) with map and reduce framework. This method takes the mean of each attribute and find the variance in each instance. The proposed model is evaluated and analyzed by the accuracy performance and time complexity. This research utilized various classifier like artificial neural network (ANN), support vector machine (SVM), random forest (RF), k-nearest neighbors (KNN), and AdaBoost to measure the classification performance on the big data. Proposed HBRFE model is compared with different feature selection like RFE, relief, backwards feature elimination, maximum relevance k-nearest neighbors (MR-KNN), and scalable deep ensemble framework big data classification (SDELF-BDC).
14.
Dubey, Pushpalata; Vasumathi, A.; Shwetha, B.; Suchitha, M.; Rani, T.
Railway track inspection and fault detection using autonomous robotic vehicles Journal Article
In: vol. 3257, pp. 020048+, 2025, ISBN: 0094-243X.
@article{601,
title = {Railway track inspection and fault detection using autonomous robotic vehicles},
author = {Pushpalata Dubey and A. Vasumathi and B. Shwetha and M. Suchitha and T. Rani},
url = {https://pubs.aip.org/aip/acp/article-abstract/3257/1/020048/3351484/Railway-track-inspection-and-fault-detection-using?redirectedFrom=fulltext},
isbn = {0094-243X},
year = {2025},
date = {2025-01-01},
volume = {3257},
pages = {020048+},
abstract = {In this study, an improved method for inspection of railway track crack detection is proposed. Because safety maintenance and inspection are crucial for ensuring efficient operations. Traditional methods of inspection often rely on manual labour, which can be time-consuming, costly, and prone to human error. To address these challenges, this paper proposes an innovative approach leveraging autonomous robotic vehicles and deep learning techniques for railway track inspection and fault detection. An infrared (IR) sensor, ultrasonic sensor, acoustic sensor, microcontroller and also LiDAR sensor are used for a wide range of detection. The initial focus was on six specific types of track faults: wheel burns, loose nuts and bolts, damaged sleepers, track creep, low joints, and issues with points and crossings. The extracted features are fed into a deep learning neural network to distinguish between cracked and non-cracked track images. Expected Results are obtained with the help of image processing and convolutional neural network and improved YOLOv5 algorithm with accuracy of 92.9% and an error rate of 1.5%. The railway inspection system uses a unique combination of autonomous robotic vehicles and deep learning, setting it apart from traditional methods.},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
In this study, an improved method for inspection of railway track crack detection is proposed. Because safety maintenance and inspection are crucial for ensuring efficient operations. Traditional methods of inspection often rely on manual labour, which can be time-consuming, costly, and prone to human error. To address these challenges, this paper proposes an innovative approach leveraging autonomous robotic vehicles and deep learning techniques for railway track inspection and fault detection. An infrared (IR) sensor, ultrasonic sensor, acoustic sensor, microcontroller and also LiDAR sensor are used for a wide range of detection. The initial focus was on six specific types of track faults: wheel burns, loose nuts and bolts, damaged sleepers, track creep, low joints, and issues with points and crossings. The extracted features are fed into a deep learning neural network to distinguish between cracked and non-cracked track images. Expected Results are obtained with the help of image processing and convolutional neural network and improved YOLOv5 algorithm with accuracy of 92.9% and an error rate of 1.5%. The railway inspection system uses a unique combination of autonomous robotic vehicles and deep learning, setting it apart from traditional methods.
15.
V, Kesavan M; Kumar, Josephine; Ashwin, Nanda
A novel scalable deep ensemble learning framework for big data classification via MapReduce integration Journal Article
In: vol. 14, pp. 1386+, 2025.
@article{602,
title = {A novel scalable deep ensemble learning framework for big data classification via MapReduce integration},
author = {Kesavan M V and Josephine Kumar and Nanda Ashwin},
url = {https://ijai.iaescore.com/index.php/IJAI/article/view/25483},
doi = {10.11591/ijai.v14.i2.pp1386-1400},
year = {2025},
date = {2025-01-01},
volume = {14},
pages = {1386+},
abstract = {Big data classification involves the systematic sorting and analysis of extensive datasets that are aggregated from a variety of sources. These datasets may include but are not limited to, electronic records, digital imaging, genetic information sequences, transactional data, research outputs, and data streams from wearable technologies and connected devices. This paper introduces the scalable deep ensemble learning framework for big data classification (SDELF-BDC), a novel methodology tailored for the classification of large-scale data. At its core, SDELF-BDC leverages a Hadoop-based map-reduce framework for feature selection, significantly reducing feature-length and enhancing computational efficiency. The methodology is further augmented by a deep ensemble model that judiciously applies a variety of deep learning classifiers based on data characteristics, thereby ensuring optimal performance. Each classifier's output undergoes a rigorous optimization-based ensemble approach for refinement, utilizing a sophisticated algorithm. The result is a robust classification system that excels in predictive accuracy while maintaining scalability and responsiveness to the dynamic requirements of big data environments. Through a strategic combination of classifiers and an innovative reduction phase, SDELF-BDC emerges as a comprehensive solution for big data classification challenges, setting new benchmarks for predictive analytics in diverse and data-intensive domains.},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Big data classification involves the systematic sorting and analysis of extensive datasets that are aggregated from a variety of sources. These datasets may include but are not limited to, electronic records, digital imaging, genetic information sequences, transactional data, research outputs, and data streams from wearable technologies and connected devices. This paper introduces the scalable deep ensemble learning framework for big data classification (SDELF-BDC), a novel methodology tailored for the classification of large-scale data. At its core, SDELF-BDC leverages a Hadoop-based map-reduce framework for feature selection, significantly reducing feature-length and enhancing computational efficiency. The methodology is further augmented by a deep ensemble model that judiciously applies a variety of deep learning classifiers based on data characteristics, thereby ensuring optimal performance. Each classifier's output undergoes a rigorous optimization-based ensemble approach for refinement, utilizing a sophisticated algorithm. The result is a robust classification system that excels in predictive accuracy while maintaining scalability and responsiveness to the dynamic requirements of big data environments. Through a strategic combination of classifiers and an innovative reduction phase, SDELF-BDC emerges as a comprehensive solution for big data classification challenges, setting new benchmarks for predictive analytics in diverse and data-intensive domains.
16.
S, G.; S, S.; Basavaraja, V.; Naikodi, C.; Srinivas, B.; Rao, B.
IoT-Based Home Automation and Monitoring System using ESP01 and ThingSpeak Proceedings
2025.
@proceedings{603,
title = {IoT-Based Home Automation and Monitoring System using ESP01 and ThingSpeak},
author = {G. S and S. S and V. Basavaraja and C. Naikodi and B. Srinivas and B. Rao},
url = {https://ieeexplore.ieee.org/document/11070612},
doi = {10.1109/OTCON65728.2025.11070612},
year = {2025},
date = {2025-01-01},
journal = {2025 4th OPJU International Technology Conference (OTCON) on Smart Computing for Innovation and Advancement in Industry 5.0},
pages = {1-6,},
abstract = {This study delves into the development of an IoT-based home automation and monitoring system, utilizing the Arduino Uno microcontroller in a simulated environment. Addressing the increasing demand for smart home solutions, this research highlights the significance of energy efficiency, security, and convenience in contemporary living. The system encompasses a variety of sensors, including those for temperature, light, motion, gas, and ultrasonic measurements, all linked to the Arduino Uno. Data collected by these sensors is transmitted to the cloud-based ThingSpeak platform, enabling remote monitoring and control. Major accomplishments of this project include the effective management of temperature and lighting, improved security through motion and gas detection, remote surveillance capabilities, and the successful application of simulation techniques for the rapid development of IoT solutions.},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
This study delves into the development of an IoT-based home automation and monitoring system, utilizing the Arduino Uno microcontroller in a simulated environment. Addressing the increasing demand for smart home solutions, this research highlights the significance of energy efficiency, security, and convenience in contemporary living. The system encompasses a variety of sensors, including those for temperature, light, motion, gas, and ultrasonic measurements, all linked to the Arduino Uno. Data collected by these sensors is transmitted to the cloud-based ThingSpeak platform, enabling remote monitoring and control. Major accomplishments of this project include the effective management of temperature and lighting, improved security through motion and gas detection, remote surveillance capabilities, and the successful application of simulation techniques for the rapid development of IoT solutions.
17.
S, Shiva; S, Ganga; Basavaraja, Venkappanavara; Naikodi, Dr. Chandrakant; Srinivas, Badrinath; Rao, B.
2025.
@book{604,
title = {A Novel Approach to Authenticate Smartphone using Sequence of Multiple Fingerprints within Blockchain Network},
author = {Shiva S and Ganga S and Venkappanavara Basavaraja and Dr. Chandrakant Naikodi and Badrinath Srinivas and B. Rao},
url = {https://ieeexplore.ieee.org/document/11070870},
doi = {10.1109/OTCON65728.2025.11070870},
year = {2025},
date = {2025-01-01},
pages = {1-5,},
abstract = {In recent times biometric and blockchain are the two main key terms using for security and authentication purpose. In this paper we are proposing a novel approach to ensure the security, integrity, and traceability of sensitive data with integration of fingerprint and sequence ID stored within blockchain network. Without depending on a single authority, the decentralized and unchangeable nature of blockchain provides a strong framework for safely storing and confirming identityrelated data. This work investigates the use of cryptographic hash functions to sequence IDs and fingerprint data representation on the blockchain, preserving user privacy and guaranteeing data integrity with the help of blockchain. It also outlines the possible drawbacks of blockchain, including scalability, high resource consumption, and privacy concerns, in addition to its benefits, which include improved security, tamper-proof records, and transparent data access. Blockchain technology combined with fingerprint and sequence data storage offers a viable solution for high-trust applications including identity management, safe data sharing, and forensic investigations.},
keywords = {CSE},
pubstate = {published},
tppubtype = {book}
}
In recent times biometric and blockchain are the two main key terms using for security and authentication purpose. In this paper we are proposing a novel approach to ensure the security, integrity, and traceability of sensitive data with integration of fingerprint and sequence ID stored within blockchain network. Without depending on a single authority, the decentralized and unchangeable nature of blockchain provides a strong framework for safely storing and confirming identityrelated data. This work investigates the use of cryptographic hash functions to sequence IDs and fingerprint data representation on the blockchain, preserving user privacy and guaranteeing data integrity with the help of blockchain. It also outlines the possible drawbacks of blockchain, including scalability, high resource consumption, and privacy concerns, in addition to its benefits, which include improved security, tamper-proof records, and transparent data access. Blockchain technology combined with fingerprint and sequence data storage offers a viable solution for high-trust applications including identity management, safe data sharing, and forensic investigations.
18.
Asokaraj, Manimegalai; Kumar, Josephine; Ashwin, Nanda
An improved similarity matching model for the content-based image retrieval model Journal Article
In: vol. 18, 2025.
@article{605,
title = {An improved similarity matching model for the content-based image retrieval model},
author = {Manimegalai Asokaraj and Josephine Kumar and Nanda Ashwin},
url = {https://sciendo.com/article/10.2478/ijssis-2025-0031},
doi = {10.2478/ijssis-2025-0031},
year = {2025},
date = {2025-01-01},
volume = {18},
abstract = {Content-based retrieval (CBR) is an essential process to retrieve images from databases based on metadata from the image. Metadata in an image refers to image colors, textures, and shapes, or any other important information that can be derived from the image itself. The goal of CBR is to search for the relevant image and retrieve it from databases. Many CBR models achieved reliable results in analyzing the image. However, the computation cost of image retrieval is a challenging task due to the growth of image traits. This paper presents an Optimized Hybrid Ensemble Model (OHEM) R-2,C-1. The proposed model aims to improve the process of similarity matching for the efficient analysis of the query image while minimizing computation time. The purpose of OHEM is twofold. First, OHEM analyzes the features within the query image and performs similarity matching within the database, achieving this with reduced computational complexity. Subsequently, in accordance with the established objective function, it identifies and evaluates the pertinent features. Two distinct datasets, ROxford and RParis, are utilized to assess the model's performance. Several assessment criteria, including F1-score, recall, precision, and computation time, have been used to assess the model. The computation and evaluated outcomes are compared to six distinct algorithms, such as CSM, equilibrium propagation (EP), DCM, GA-based IR, and IRT. The comparison findings suggested that the proposed method performs better than the other models. R-2,C-1.},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Content-based retrieval (CBR) is an essential process to retrieve images from databases based on metadata from the image. Metadata in an image refers to image colors, textures, and shapes, or any other important information that can be derived from the image itself. The goal of CBR is to search for the relevant image and retrieve it from databases. Many CBR models achieved reliable results in analyzing the image. However, the computation cost of image retrieval is a challenging task due to the growth of image traits. This paper presents an Optimized Hybrid Ensemble Model (OHEM) R-2,C-1. The proposed model aims to improve the process of similarity matching for the efficient analysis of the query image while minimizing computation time. The purpose of OHEM is twofold. First, OHEM analyzes the features within the query image and performs similarity matching within the database, achieving this with reduced computational complexity. Subsequently, in accordance with the established objective function, it identifies and evaluates the pertinent features. Two distinct datasets, ROxford and RParis, are utilized to assess the model's performance. Several assessment criteria, including F1-score, recall, precision, and computation time, have been used to assess the model. The computation and evaluated outcomes are compared to six distinct algorithms, such as CSM, equilibrium propagation (EP), DCM, GA-based IR, and IRT. The comparison findings suggested that the proposed method performs better than the other models. R-2,C-1.
19.
AkshayKumar, B. T.; Harshith, R.; Preetham, S. U.; Priyadarshini, M.
Brain Tumor Classification and Detection with VGG-16 using MRI Images Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331531034 (ISBN), (0).
@proceedings{638,
title = {Brain Tumor Classification and Detection with VGG-16 using MRI Images},
author = {B. T. AkshayKumar and R. Harshith and S. U. Preetham and M. Priyadarshini},
url = {https://ieeexplore.ieee.org/document/11139855},
doi = {10.1109/INCET64471.2025.11139855},
isbn = {9798331531034 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {To improve the precision and treatment, classification and detection of brain tumors is important. In this study for classification and detection of brain tumors from MRI images, the VGG16[18] convolutional neural network (CNN) is used. In this study, the dataset consists of labeled MRI images[11] of patients having tumors and not having tumors. The proposed approach employs transfer learning with a pre-trained VGG16 network for feature extraction and fine-tuning for binary classification (tumor/no tumor). Scaling, normalization, and augmentation are examples of picture preprocessing methods used to increase dataset diversity. With an overall accuracy of 95.78% and an F1-score of 95.17% on the test set, the model proved to be successful in distinguishing between tumorous and non-tumorous areas. These promising results suggest that the VGG16-based approach can support improved clinical judgment by assisting in the timely and accurate diagnosis of brain tumors. Other CNN architectures[17] will be investigated in future research, and the dataset will be expanded for better generalization.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
To improve the precision and treatment, classification and detection of brain tumors is important. In this study for classification and detection of brain tumors from MRI images, the VGG16[18] convolutional neural network (CNN) is used. In this study, the dataset consists of labeled MRI images[11] of patients having tumors and not having tumors. The proposed approach employs transfer learning with a pre-trained VGG16 network for feature extraction and fine-tuning for binary classification (tumor/no tumor). Scaling, normalization, and augmentation are examples of picture preprocessing methods used to increase dataset diversity. With an overall accuracy of 95.78% and an F1-score of 95.17% on the test set, the model proved to be successful in distinguishing between tumorous and non-tumorous areas. These promising results suggest that the VGG16-based approach can support improved clinical judgment by assisting in the timely and accurate diagnosis of brain tumors. Other CNN architectures[17] will be investigated in future research, and the dataset will be expanded for better generalization.
20.
Kiran, M.; Hariharan, S.; Akshay, K. M.; Leangashwar, G.
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331531034 (ISBN), (0).
@proceedings{639,
title = {Data Security and Protection: A Mechanism for Managing Data Theft and Cybercrime in Online Platforms of Educational Institution},
author = {M. Kiran and S. Hariharan and K. M. Akshay and G. Leangashwar},
url = {https://ieeexplore.ieee.org/document/11140967},
doi = {10.1109/INCET64471.2025.11140967},
isbn = {9798331531034 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Phishing attacks, which use fake URLs to imitate trustworthy websites, are still a constant and developing cybersecurity risk. The machine learning method for automated phishing URL detection is examined in this research. We introduce a system that uses a Random Forest Classifier that was learned using features that were taken straight out of URL strings. The length of the URL, whether the 'https' protocol indicator is present, and the number of particular special characters are some examples of these characteristics. Our system, which is implemented in Python and uses tools like Flask for API deployment, pandas for data manipulation, and scikit-learn for model creation, shows a workable and effective approach. On a dataset of phishing URLs, experimental testing shows that the trained Random Forest Classifier attains a classification accuracy of 0.89.The effectiveness of using straightforward, computationally cheap URL-based characteristics in combination with ensemble learning approaches for successful preliminary phishing detection is demonstrated by this finding. The developed system provides an easily integrable component for improved web security applications and real-time URL analysis, and it is available through a RESTful API.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Phishing attacks, which use fake URLs to imitate trustworthy websites, are still a constant and developing cybersecurity risk. The machine learning method for automated phishing URL detection is examined in this research. We introduce a system that uses a Random Forest Classifier that was learned using features that were taken straight out of URL strings. The length of the URL, whether the 'https' protocol indicator is present, and the number of particular special characters are some examples of these characteristics. Our system, which is implemented in Python and uses tools like Flask for API deployment, pandas for data manipulation, and scikit-learn for model creation, shows a workable and effective approach. On a dataset of phishing URLs, experimental testing shows that the trained Random Forest Classifier attains a classification accuracy of 0.89.The effectiveness of using straightforward, computationally cheap URL-based characteristics in combination with ensemble learning approaches for successful preliminary phishing detection is demonstrated by this finding. The developed system provides an easily integrable component for improved web security applications and real-time URL analysis, and it is available through a RESTful API.
21.
Shilpa, V.; Girija, V.; Srimithun, G.; Sara, Nishita; Sneha, K.
Ransomware Detection and Prevention Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331531034 (ISBN), (0).
@proceedings{640,
title = {Ransomware Detection and Prevention},
author = {V. Shilpa and V. Girija and G. Srimithun and Nishita Sara and K. Sneha},
url = {https://ieeexplore.ieee.org/document/11140093},
doi = {10.1109/INCET64471.2025.11140093},
isbn = {9798331531034 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Ransomware is a rapidly evolving cyber threat that encrypts user data and demands a ransom for its release, causing severe financial and operational disruptions. The following study integrates behavioral study, real-time file scrutiny, and signature- based methods into one comprehensive approach to combat ransomware attacks. The proposed system employs heuristic methods and machine learning techniques to identify unwanted accesses, encryption attempts, and other suspicious file activity. Furthermore, the method is anti- ransom proactive in that damage containment through process isolation and file backup is executed prior to the successful deployment of the ransomware. An interactive interface is presented in which infected files can be quarantined by checking them for malware and compromised data can be retrieved from safe storage. The results indicate that it is capable of mounting further regime against the ransomware as well as protecting the afflicted user or group, preserving the data, and saving a lot of money that may go into financial suffering.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Ransomware is a rapidly evolving cyber threat that encrypts user data and demands a ransom for its release, causing severe financial and operational disruptions. The following study integrates behavioral study, real-time file scrutiny, and signature- based methods into one comprehensive approach to combat ransomware attacks. The proposed system employs heuristic methods and machine learning techniques to identify unwanted accesses, encryption attempts, and other suspicious file activity. Furthermore, the method is anti- ransom proactive in that damage containment through process isolation and file backup is executed prior to the successful deployment of the ransomware. An interactive interface is presented in which infected files can be quarantined by checking them for malware and compromised data can be retrieved from safe storage. The results indicate that it is capable of mounting further regime against the ransomware as well as protecting the afflicted user or group, preserving the data, and saving a lot of money that may go into financial suffering.
22.
Venunadh, A.; Aparna, N.
NOISEX:A Classifier for Music Instruments Using Convolutional Neural Network Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331531034 (ISBN), (0).
@proceedings{641,
title = {NOISEX:A Classifier for Music Instruments Using Convolutional Neural Network},
author = {A. Venunadh and N. Aparna},
url = {https://ieeexplore.ieee.org/document/11140359},
doi = {10.1109/INCET64471.2025.11140359},
isbn = {9798331531034 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {NoiseX is an extensive machine learning initiative created to tackle numerous audios processing issues, especially concentrating on the categorization of various musical instruments. This effort additionally includes various signal-processing methods to address societal problems linked to sound analysis. The initiative entails utilizing an actual dataset, investigating how machine learning techniques can reveal concealed patterns within the information. The main goal was to acquire practical experience with data-mining techniques and machine learning libraries, leading to a report that outlines the dataset and algorithms used. In the musical instrument classification project, we utilized sophisticated techniques including spectral analysis, deep neural networks (DNNs), and convolutional neural networks (CNNs) to enhance classification precision. In this way, we also investigated the possibilities of machine learning in practical uses of audio identification and categorization. This procedure not only improved our technical knowledge but also underscored the difficulties in precisely processing and categorizing intricate audio signals.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
NoiseX is an extensive machine learning initiative created to tackle numerous audios processing issues, especially concentrating on the categorization of various musical instruments. This effort additionally includes various signal-processing methods to address societal problems linked to sound analysis. The initiative entails utilizing an actual dataset, investigating how machine learning techniques can reveal concealed patterns within the information. The main goal was to acquire practical experience with data-mining techniques and machine learning libraries, leading to a report that outlines the dataset and algorithms used. In the musical instrument classification project, we utilized sophisticated techniques including spectral analysis, deep neural networks (DNNs), and convolutional neural networks (CNNs) to enhance classification precision. In this way, we also investigated the possibilities of machine learning in practical uses of audio identification and categorization. This procedure not only improved our technical knowledge but also underscored the difficulties in precisely processing and categorizing intricate audio signals.
23.
Dev, A.; Sanjay, H.; Vishnu, V.; Priyadarshini, M.
Improving Liver Tumor Segmentation Robustness with Physics -Informed Regularization of a ResNet50 Network Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331531034 (ISBN), (0).
@proceedings{644,
title = {Improving Liver Tumor Segmentation Robustness with Physics -Informed Regularization of a ResNet50 Network},
author = {A. Dev and H. Sanjay and V. Vishnu and M. Priyadarshini},
url = {https://ieeexplore.ieee.org/document/11140218},
doi = {10.1109/INCET64471.2025.11140218},
isbn = {9798331531034 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {The clinical process of liver cancer diagnosis, treatment planning, and monitoring depends on the precise segmentation of liver tumors using 'Computed Tomography (CT) scans'. The segmentation of medical images has advanced significantly, thanks to deep learning techniques, particularly 'Convolutional Neural Networks', or CNNs. However, data-driven models by themselves might produce results that are not anatomically reasonable or that do not sufficiently account for known picture properties. This study investigates the application of Physics-Informed Neural Networks (PINNs) to enhance the segmentation performance of a ResNet50-based model in order to get beyond these limitations. By including physics-based constraints into the training process, we hope to direct the network to produce liver tumor segmentations that are more reliable and clinically significant. Our method uses a ResNet50 architecture to generate a baseline segmentation model, which is subsequently adjusted with a bespoke loss function. With a focus on encouraging uniformity in intensity inside the segmented regions and smoothness in the segmented tumor boundaries, this physics-informed loss is intended to penalize departures from expected image characteristics. Our PINN-enhanced model was tested on a specific test dataset against the baseline ResNet50. The quantitative findings show that the PINN technique effectively classifies pixels at the pixel level, achieving a high Accuracy of 97.15%, a robust F1-Score of 95.74%, and good Specificity of 100.00%. While the Dice Coefficient of 32.85% and Mean Intersection over Union (IoU) of 32.38% suggests that there could still be difficulties in obtaining precise region overlap, the low 'Mean Absolute Error (MAE)' of 0.0288 indicates that pixel predictions are generally correct.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
The clinical process of liver cancer diagnosis, treatment planning, and monitoring depends on the precise segmentation of liver tumors using 'Computed Tomography (CT) scans'. The segmentation of medical images has advanced significantly, thanks to deep learning techniques, particularly 'Convolutional Neural Networks', or CNNs. However, data-driven models by themselves might produce results that are not anatomically reasonable or that do not sufficiently account for known picture properties. This study investigates the application of Physics-Informed Neural Networks (PINNs) to enhance the segmentation performance of a ResNet50-based model in order to get beyond these limitations. By including physics-based constraints into the training process, we hope to direct the network to produce liver tumor segmentations that are more reliable and clinically significant. Our method uses a ResNet50 architecture to generate a baseline segmentation model, which is subsequently adjusted with a bespoke loss function. With a focus on encouraging uniformity in intensity inside the segmented regions and smoothness in the segmented tumor boundaries, this physics-informed loss is intended to penalize departures from expected image characteristics. Our PINN-enhanced model was tested on a specific test dataset against the baseline ResNet50. The quantitative findings show that the PINN technique effectively classifies pixels at the pixel level, achieving a high Accuracy of 97.15%, a robust F1-Score of 95.74%, and good Specificity of 100.00%. While the Dice Coefficient of 32.85% and Mean Intersection over Union (IoU) of 32.38% suggests that there could still be difficulties in obtaining precise region overlap, the low 'Mean Absolute Error (MAE)' of 0.0288 indicates that pixel predictions are generally correct.
24.
Kumar, J. P.; Shreyas, M.; Goni, V. R.; Tejas, B.
Measure Size of Objects in an Image using Computer Vision and OpenCV Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331524760 (ISBN), (0).
@proceedings{648,
title = {Measure Size of Objects in an Image using Computer Vision and OpenCV},
author = {J. P. Kumar and M. Shreyas and V. R. Goni and B. Tejas},
url = {https://ieeexplore.ieee.org/document/11108574},
doi = {10.1109/ETCC65847.2025.11108574},
isbn = {9798331524760 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Object measurement in images is crucial in computer vision, with applications in industrial automation, quality control, and medical imaging. Traditional manual methods are inefficient and error-prone, while image processing techniques improve accuracy and efficiency. This study introduces an automated measurement system using OpenCV, integrating image preprocessing, edge detection, and contour extraction. The process involves grayscale conversion, Gaussian blur for noise reduction, and Canny edge detection to define object boundaries. Contour filtering isolates objects, and a reference object establishes a pixel-to-metric ratio for precise measurements. Euclidean distance calculations determine dimensions, achieving an error rate below 5% in most cases. Additionally, graphical visualizations enhance result interpretation. This cost-effective and scalable solution benefits industries like manufacturing, logistics, and healthcare by improving measurement precision and reducing human error. By leveraging automated image processing, the system enhances efficiency, accuracy, and applicability in real-world scenarios requiring precise object measurements.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Object measurement in images is crucial in computer vision, with applications in industrial automation, quality control, and medical imaging. Traditional manual methods are inefficient and error-prone, while image processing techniques improve accuracy and efficiency. This study introduces an automated measurement system using OpenCV, integrating image preprocessing, edge detection, and contour extraction. The process involves grayscale conversion, Gaussian blur for noise reduction, and Canny edge detection to define object boundaries. Contour filtering isolates objects, and a reference object establishes a pixel-to-metric ratio for precise measurements. Euclidean distance calculations determine dimensions, achieving an error rate below 5% in most cases. Additionally, graphical visualizations enhance result interpretation. This cost-effective and scalable solution benefits industries like manufacturing, logistics, and healthcare by improving measurement precision and reducing human error. By leveraging automated image processing, the system enhances efficiency, accuracy, and applicability in real-world scenarios requiring precise object measurements.
25.
Upadhyay, A.; Joy, Alisha; Rohan, R.; Kumar, J. P.
Virtual Stylist: Outfit Try-On and Personalized Fashion Advice Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331524760 (ISBN), (0).
@proceedings{649,
title = {Virtual Stylist: Outfit Try-On and Personalized Fashion Advice},
author = {A. Upadhyay and Alisha Joy and R. Rohan and J. P. Kumar},
url = {https://ieeexplore.ieee.org/document/11108640},
doi = {10.1109/ETCC65847.2025.11108640},
isbn = {9798331524760 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {This project addresses the challenge of personalized style inspiration in today’s fast-paced fashion landscape. Many individuals struggle to visualize outfits and lack access to tailored fashion advice. This application introduces an AI-powered virtual stylist that provides mood-based outfit recommendations and a real-time virtual try-on experience using advanced computer vision and image processing. An interactive chatbot enhances engagement by offering personalized fashion advice. The system learns user preferences over time, refining recommendations for a more accurate and tailored experience. Findings highlight the effectiveness of mood-based styling, the impact of virtual try-ons on decision-making, and the chatbot’s role in enhancing user satisfaction. This project revolutionizes fashion discovery by integrating cutting-edge technology with personalized styling.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
This project addresses the challenge of personalized style inspiration in today’s fast-paced fashion landscape. Many individuals struggle to visualize outfits and lack access to tailored fashion advice. This application introduces an AI-powered virtual stylist that provides mood-based outfit recommendations and a real-time virtual try-on experience using advanced computer vision and image processing. An interactive chatbot enhances engagement by offering personalized fashion advice. The system learns user preferences over time, refining recommendations for a more accurate and tailored experience. Findings highlight the effectiveness of mood-based styling, the impact of virtual try-ons on decision-making, and the chatbot’s role in enhancing user satisfaction. This project revolutionizes fashion discovery by integrating cutting-edge technology with personalized styling.
26.
Yashaswini, S.; Kumar, K. V.; Vasumathi, A. K.
IoT device for monitoring air pollution Journal Article
In: Asian Textile Journal, vol. 34, pp. 42-47,, 2025, ISBN: 09713425 (ISSN), (0).
@article{726,
title = {IoT device for monitoring air pollution},
author = {S. Yashaswini and K. V. Kumar and A. K. Vasumathi},
url = {https://www.atjournal.com/images/november-2025.pdf},
isbn = {09713425 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Asian Textile Journal},
volume = {34},
pages = {42-47,},
publisher = {G P S Kwatra},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
27.
Prasanna, Mahesh; Jayanthi, M. G.; Balasubramani, R.; Benal, G. D.; Barve, A.; Shelke, A.
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331536770 (ISBN), (0).
@proceedings{727,
title = {Dynamic Trust-Based Authentication Framework for Mobile Users Accessing Distributed Multi-Cloud Infrastructure},
author = {Mahesh Prasanna and M. G. Jayanthi and R. Balasubramani and G. D. Benal and A. Barve and A. Shelke},
url = {https://ieeexplore.ieee.org/document/11211038},
doi = {10.1109/IACIS65746.2025.11211038},
isbn = {9798331536770 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Mobile cloud computing environments encounter considerable security challenges due to the ever-changing nature of mobile users and the decentralized architecture of multi-cloud systems. Conventional authentication methods do not adequately address the unpredictable behavior patterns and resource-limited characteristics of mobile devices that access diverse cloud services. A dynamic trust-based authentication framework has been developed, which integrates behavioral analysis, contextual awareness, and multi-factor risk assessment algorithms. This system utilizes machine learning techniques, employing bidirectional GRU models and multi-agent deep deterministic policy gradient optimization to facilitate real-time trust score calculations and adaptive access control decisions. The proposed framework has achieved an authentication accuracy of 96.2%, representing a 14.78% improvement over baseline methods, alongside a threat detection rate of 99.1% and a false positive rate of 2.3%. The average response time for trust evaluation was 127ms, with a user satisfaction rate of 94.8% in usability assessments. This framework effectively addresses security vulnerabilities while preserving user experience, showcasing a significant advancement in mobile cloud authentication with improved adaptability to dynamic user behaviors and distributed cloud environments.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Mobile cloud computing environments encounter considerable security challenges due to the ever-changing nature of mobile users and the decentralized architecture of multi-cloud systems. Conventional authentication methods do not adequately address the unpredictable behavior patterns and resource-limited characteristics of mobile devices that access diverse cloud services. A dynamic trust-based authentication framework has been developed, which integrates behavioral analysis, contextual awareness, and multi-factor risk assessment algorithms. This system utilizes machine learning techniques, employing bidirectional GRU models and multi-agent deep deterministic policy gradient optimization to facilitate real-time trust score calculations and adaptive access control decisions. The proposed framework has achieved an authentication accuracy of 96.2%, representing a 14.78% improvement over baseline methods, alongside a threat detection rate of 99.1% and a false positive rate of 2.3%. The average response time for trust evaluation was 127ms, with a user satisfaction rate of 94.8% in usability assessments. This framework effectively addresses security vulnerabilities while preserving user experience, showcasing a significant advancement in mobile cloud authentication with improved adaptability to dynamic user behaviors and distributed cloud environments.
28.
Kanan, M.; Muraliraja, R.; Ramanathan, V.
Performance optimization of CI engine using python with blends of waste oil biodiesel, plastic pyrolysis oil, and diesel Journal Article
In: Results in Chemistry, vol. 18, 2025, (0).
@article{718,
title = {Performance optimization of CI engine using python with blends of waste oil biodiesel, plastic pyrolysis oil, and diesel},
author = {M. Kanan and R. Muraliraja and V. Ramanathan},
url = {https://www.sciencedirect.com/science/article/pii/S2211715625007532?via%3Dihub},
doi = {10.1016/j.rechem.2025.102769},
year = {2025},
date = {2025-01-01},
journal = {Results in Chemistry},
volume = {18},
publisher = {Elsevier B.V.},
abstract = {This study explores the novel integration of ternary fuel blends comprising diesel, biodiesel derived from waste cooking oil, and pyrolysis oil obtained from waste plastics in a single-cylinder direct injection compression ignition (CI) engine. A major innovation lies in the simultaneous utilization of two waste-derived fuels—biodiesel and plastic pyrolysis oil—to create a sustainable, low-emission alternative to conventional diesel. Engine tests were conducted under varying injection pressures (170, 200, and 230 bar) and load conditions to evaluate brake thermal efficiency (BTE), exhaust gas temperature (EGT), and key emissions such as NOx, CO, UHC, and smoke opacity. The results indicate that blends such as P10B220D70 and P15B215D70 offer optimal engine performance and combustion stability at 230 bar injection pressure. Notably, a Python-based statistical correlation analysis was employed to determine the influence of input variables (fuel blend, injection pressure, and load) on engine performance and emissions, identifying injection pressure as the most significant factor. This integrated experimental and computational approach underscores the viability of ternary waste-fuel blends as a promising solution for cleaner and more efficient CI engine operation.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
This study explores the novel integration of ternary fuel blends comprising diesel, biodiesel derived from waste cooking oil, and pyrolysis oil obtained from waste plastics in a single-cylinder direct injection compression ignition (CI) engine. A major innovation lies in the simultaneous utilization of two waste-derived fuels—biodiesel and plastic pyrolysis oil—to create a sustainable, low-emission alternative to conventional diesel. Engine tests were conducted under varying injection pressures (170, 200, and 230 bar) and load conditions to evaluate brake thermal efficiency (BTE), exhaust gas temperature (EGT), and key emissions such as NOx, CO, UHC, and smoke opacity. The results indicate that blends such as P10B220D70 and P15B215D70 offer optimal engine performance and combustion stability at 230 bar injection pressure. Notably, a Python-based statistical correlation analysis was employed to determine the influence of input variables (fuel blend, injection pressure, and load) on engine performance and emissions, identifying injection pressure as the most significant factor. This integrated experimental and computational approach underscores the viability of ternary waste-fuel blends as a promising solution for cleaner and more efficient CI engine operation.
29.
Harshvardhan, A.; Shaikh, R. J.; Khan, K. V.; Jayanthi, M. G.
In: Sensing and Imaging, vol. 26, 2025, ISBN: 15572064 (ISSN), (0).
@article{716,
title = {Adaptive Dual-Channel Neural Network with Triangulation Topology Optimization for Kidney Cancer Diagnosis and Surgery Planning Using Clinical Metadata},
author = {A. Harshvardhan and R. J. Shaikh and K. V. Khan and M. G. Jayanthi},
url = {https://link.springer.com/article/10.1007/s11220-025-00674-7},
doi = {10.1007/s11220-025-00674-7},
isbn = {15572064 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Sensing and Imaging},
volume = {26},
publisher = {Springer},
abstract = {Heterogeneity in tumor size, kind, and stage makes it difficult to diagnose kidney cancer and prepare for surgery, making the decision between partial & radical nephrectomy more difficult. This study proposes an Adaptive Dual-Channel Pulse-Coupled Neural Network with Triangulation Topology Aggregation Optimizer (ADP-CNN-TTAO) that integrates computed tomography (CT) imaging and clinical metadata for more reliable decision support. Using the publicly available KiTS21 dataset comprising 300 annotated patient cases with diverse tumor subtypes, the method combines Iterative Robust Peak-Aware Guided Filtering (IRPAGF) for CT preprocessing, robust imputation for missing clinical variables, Analytical Clifford Fourier Mellin Transform (ACFMT) for imaging feature extraction, and Steerable Transformers (ST) for metadata representation. Experimental evaluation with cross–validation shows consistently high classification performance across papillary, chromophobe, clear cell, and oncocytoma subtypes, outperforming state-of-the-art baselines. Importantly, tumor volume and stage emerged as key determinants for surgical planning. While results demonstrate strong potential for clinical decision support, the approach requires further validation on multi-center datasets and real-world prospective trials to confirm its generalizability and clinical impact.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Heterogeneity in tumor size, kind, and stage makes it difficult to diagnose kidney cancer and prepare for surgery, making the decision between partial & radical nephrectomy more difficult. This study proposes an Adaptive Dual-Channel Pulse-Coupled Neural Network with Triangulation Topology Aggregation Optimizer (ADP-CNN-TTAO) that integrates computed tomography (CT) imaging and clinical metadata for more reliable decision support. Using the publicly available KiTS21 dataset comprising 300 annotated patient cases with diverse tumor subtypes, the method combines Iterative Robust Peak-Aware Guided Filtering (IRPAGF) for CT preprocessing, robust imputation for missing clinical variables, Analytical Clifford Fourier Mellin Transform (ACFMT) for imaging feature extraction, and Steerable Transformers (ST) for metadata representation. Experimental evaluation with cross–validation shows consistently high classification performance across papillary, chromophobe, clear cell, and oncocytoma subtypes, outperforming state-of-the-art baselines. Importantly, tumor volume and stage emerged as key determinants for surgical planning. While results demonstrate strong potential for clinical decision support, the approach requires further validation on multi-center datasets and real-world prospective trials to confirm its generalizability and clinical impact.
30.
Reddy, A.; Varalakshmi, K. V.; Prasad, V. N.; Mythili, M.; Sudha, V.; Rakesh, V. S.
Automata-Based Model for SQL Injection Pattern Recognition Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331513085 (ISBN), (0).
@proceedings{704,
title = {Automata-Based Model for SQL Injection Pattern Recognition},
author = {A. Reddy and K. V. Varalakshmi and V. N. Prasad and M. Mythili and V. Sudha and V. S. Rakesh},
url = {https://ieeexplore.ieee.org/document/11188186},
doi = {10.1109/NMITCON65824.2025.11188186},
isbn = {9798331513085 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {SQL injection remains one of the most critical web application vulnerabilities, often used to bypass authentication and extract sensitive data. While many modern detection systems rely on machine learning or pattern matching, this study revisits classical computational models for recognizing interpretable SQLi patterns. Specifically, three automata-Deterministic Finite Automaton (DFA), Pushdown Automaton (PDA), and Turing Machine (TM)-are manually constructed to detect both a classic login bypass and a UNIONbased SQL injection pattern. A set of 40 handcrafted inputs was used to evaluate their pattern recognition capabilities. DFA effectively handles simple input sequences but fails with nested or logic-based constructs. PDA improves detection through stack-based handling of structured patterns, while TM provides the most comprehensive recognition by simulating conditional logic and multi-clause sequences. TM achieved 100% accuracy with no false negatives in complex cases, outperforming DFA and PDA. Although not deployable, these models highlight the pedagogical and conceptual utility of formal language theory in modeling and understanding injection attacks.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
SQL injection remains one of the most critical web application vulnerabilities, often used to bypass authentication and extract sensitive data. While many modern detection systems rely on machine learning or pattern matching, this study revisits classical computational models for recognizing interpretable SQLi patterns. Specifically, three automata-Deterministic Finite Automaton (DFA), Pushdown Automaton (PDA), and Turing Machine (TM)-are manually constructed to detect both a classic login bypass and a UNIONbased SQL injection pattern. A set of 40 handcrafted inputs was used to evaluate their pattern recognition capabilities. DFA effectively handles simple input sequences but fails with nested or logic-based constructs. PDA improves detection through stack-based handling of structured patterns, while TM provides the most comprehensive recognition by simulating conditional logic and multi-clause sequences. TM achieved 100% accuracy with no false negatives in complex cases, outperforming DFA and PDA. Although not deployable, these models highlight the pedagogical and conceptual utility of formal language theory in modeling and understanding injection attacks.
31.
Rakesh, V. S.; Vasanthakumar, G. U.
Behavioral Traffic Monitoring and Analysis in Software-Defined Networks Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331527983 (ISBN), (0).
@proceedings{680,
title = {Behavioral Traffic Monitoring and Analysis in Software-Defined Networks},
author = {V. S. Rakesh and G. U. Vasanthakumar},
url = {https://ieeexplore.ieee.org/document/11158942},
doi = {10.1109/ICCTDC64446.2025.11158942},
isbn = {9798331527983 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {In Software-Defined Networking (SDN), the centralized control model introduces both operational flexibility and new security challenges. While much research has focused on detection-based solutions, this paper presents a monitoring-centric framework for analyzing traffic behavior in SDN environments without relying on classification models. Using both the InSDN public dataset and a custom-generated dataset in a Mininet-Ryu testbed, traffic features such as flow duration, destination port entropy, and controller packet-in rates were examined. Results demonstrate that behavioral anomalies, such as entropy drops and packet-in rate spikes, can be effectively identified through statistical and control-plane analysis. The findings validate the feasibility of behavior-based traffic monitoring as a foundation for proactive network management.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
In Software-Defined Networking (SDN), the centralized control model introduces both operational flexibility and new security challenges. While much research has focused on detection-based solutions, this paper presents a monitoring-centric framework for analyzing traffic behavior in SDN environments without relying on classification models. Using both the InSDN public dataset and a custom-generated dataset in a Mininet-Ryu testbed, traffic features such as flow duration, destination port entropy, and controller packet-in rates were examined. Results demonstrate that behavioral anomalies, such as entropy drops and packet-in rate spikes, can be effectively identified through statistical and control-plane analysis. The findings validate the feasibility of behavior-based traffic monitoring as a foundation for proactive network management.
32.
Manjunath, S.; Sivasubramanian, R.; Rajalakshmi, C. N.; Kumar, S.; Ramachandra, R.
Attention-Based Spatio-Temporal Graph Neural Network for Multi-Pollutant Urban Air Quality Prediction Proceedings
Institute of Electrical and Electronics Engineers Inc., 2025, ISBN: 9798331536794 (ISBN), (0).
@proceedings{679,
title = {Attention-Based Spatio-Temporal Graph Neural Network for Multi-Pollutant Urban Air Quality Prediction},
author = {S. Manjunath and R. Sivasubramanian and C. N. Rajalakshmi and S. Kumar and R. Ramachandra},
url = {https://ieeexplore.ieee.org/document/11168500},
doi = {10.1109/ICDSNS65743.2025.11168500},
isbn = {9798331536794 (ISBN)},
year = {2025},
date = {2025-01-01},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {In modern times, air quality prediction become an essential component of smart city planning and environmental observation because of increasing urban pollution levels. However, existing Gated Recurrent Unit (GRU)-based spatiotemporal approaches face challenges such as restricted spatial adaptability, lack of attention mechanisms, and high computational cost. To address these challenges, an attention-based spatiotemporal graph convolutional network (ASTGCN) is proposed for accurate multi-pollutant air quality forecasting. Initially, multivariate time-series data were collected from Global Urban Air Quality Index Dataset. Furthermore, linear interpolation was used for missing value imputation, and min-max normalization with sliding window segmentation was used to prepare temporally aligned inputs. Then, a city-level graph is constructed using geographical proximity, where each node represents a city, and edge encodes spatial relations. Then, data fed into ASTGCN model, where graph convolution layers extract spatial features and temporal Convolutional Neural Network (CNN) layers will identify crucial time steps. Furthermore, a random search was used to tune hyperparameters, which enabled model to achieve improved generalization while reducing training cost. Finally, experimental results demonstrated that ASTGCN achieved an improved Mean Absolute Error (MAE) of 13.1%, Root Mean Squared Error (RMSE) of 19.5, and R2 of 95%.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
In modern times, air quality prediction become an essential component of smart city planning and environmental observation because of increasing urban pollution levels. However, existing Gated Recurrent Unit (GRU)-based spatiotemporal approaches face challenges such as restricted spatial adaptability, lack of attention mechanisms, and high computational cost. To address these challenges, an attention-based spatiotemporal graph convolutional network (ASTGCN) is proposed for accurate multi-pollutant air quality forecasting. Initially, multivariate time-series data were collected from Global Urban Air Quality Index Dataset. Furthermore, linear interpolation was used for missing value imputation, and min-max normalization with sliding window segmentation was used to prepare temporally aligned inputs. Then, a city-level graph is constructed using geographical proximity, where each node represents a city, and edge encodes spatial relations. Then, data fed into ASTGCN model, where graph convolution layers extract spatial features and temporal Convolutional Neural Network (CNN) layers will identify crucial time steps. Furthermore, a random search was used to tune hyperparameters, which enabled model to achieve improved generalization while reducing training cost. Finally, experimental results demonstrated that ASTGCN achieved an improved Mean Absolute Error (MAE) of 13.1%, Root Mean Squared Error (RMSE) of 19.5, and R2 of 95%.
33.
Prabhu, S.; Bharadwaj, D. Y.
Precision agriculture takes flight: Drone technology in crop management Book Chapter
In: pp. 42-64,, CRC Press, 2025, ISBN: 9781040394908 (ISBN); 9781040394939 (ISBN), (0).
@inbook{673,
title = {Precision agriculture takes flight: Drone technology in crop management},
author = {S. Prabhu and D. Y. Bharadwaj},
url = {https://www.taylorfrancis.com/chapters/edit/10.1201/9781003481584-2/precision-agriculture-takes-flight-srilakshmi-prabhu-dhanya-bharadwaj},
doi = {10.1201/9781003481584-2},
isbn = {9781040394908 (ISBN); 9781040394939 (ISBN)},
year = {2025},
date = {2025-01-01},
pages = {42-64,},
publisher = {CRC Press},
abstract = {The integration of drones, or Unmanned Aerial Systems (UAS), into agricultural practices has ushered in a new era of precision farming. Drones equipped with advanced technologies such as sensors, cameras, and data analytics offer a range of scientific and accurate benefits that revolutionize traditional agricultural methods. This chapter explores the various applications and advantages of drones in agriculture, ranging from precision agriculture and crop monitoring to pest management and yield prediction. Two case studies demonstrate the practical implementation of drone technology in detecting and managing crop pests and diseases. The current status of drone usage in Indian agriculture and the challenges hindering widespread adoption are discussed. Finally, the chapter outlines future prospects, highlighting the potential of drones and artificial intelligence (AI) in transforming agricultural practices to enhance productivity, sustainability, and food security.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {inbook}
}
The integration of drones, or Unmanned Aerial Systems (UAS), into agricultural practices has ushered in a new era of precision farming. Drones equipped with advanced technologies such as sensors, cameras, and data analytics offer a range of scientific and accurate benefits that revolutionize traditional agricultural methods. This chapter explores the various applications and advantages of drones in agriculture, ranging from precision agriculture and crop monitoring to pest management and yield prediction. Two case studies demonstrate the practical implementation of drone technology in detecting and managing crop pests and diseases. The current status of drone usage in Indian agriculture and the challenges hindering widespread adoption are discussed. Finally, the chapter outlines future prospects, highlighting the potential of drones and artificial intelligence (AI) in transforming agricultural practices to enhance productivity, sustainability, and food security.
34.
Kumar, J. P.
A Framework for Enhanced Crop and Fertilizer Recommendations Using Machine Learning, Explainable AI, and RPA Journal Article
In: International Journal of Intelligent Engineering and Systems, vol. 18, pp. 262-274,, 2025, ISBN: 21853118 (ISSN); 2185310X (ISSN), (0).
@article{672,
title = {A Framework for Enhanced Crop and Fertilizer Recommendations Using Machine Learning, Explainable AI, and RPA},
author = {J. P. Kumar},
url = {https://inass.org/wp-content/uploads/2025/07/2025113017.pdf},
doi = {10.22266/ijies2025.1130.17},
isbn = {21853118 (ISSN); 2185310X (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {International Journal of Intelligent Engineering and Systems},
volume = {18},
pages = {262-274,},
publisher = {Intelligent Network and Systems Society},
abstract = {This work explores the challenge of crop and fertilizer prediction by leveraging machine learning and AI to enhance agricultural practices, particularly for small-scale farmers in India. A major contribution in this research is the use of Robotic Process Automation (RPA) based on the UiPath platform, which enhances the accuracy and efficiency of data collection. The basis of this research puts forward a new ensemble technique that combines XGBoost and Random Forest, which demonstrates better predictive accuracy than traditional models and other ensemble techniques. The proposed framework achieved an accuracy of 98.0% and an F1 score of 98.2% on crop and fertilizer recommendation tasks, outperforming baseline models by over 3%. These results demonstrate the model’s effectiveness and reliability in real-world agricultural scenarios. We present an integrated framework that combines these ensemble techniques with Explainable AI (XAI). This technique ensures that the generated predictions are explainable, thus making it possible for stakeholders to understand and accept the recommendations offered.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
This work explores the challenge of crop and fertilizer prediction by leveraging machine learning and AI to enhance agricultural practices, particularly for small-scale farmers in India. A major contribution in this research is the use of Robotic Process Automation (RPA) based on the UiPath platform, which enhances the accuracy and efficiency of data collection. The basis of this research puts forward a new ensemble technique that combines XGBoost and Random Forest, which demonstrates better predictive accuracy than traditional models and other ensemble techniques. The proposed framework achieved an accuracy of 98.0% and an F1 score of 98.2% on crop and fertilizer recommendation tasks, outperforming baseline models by over 3%. These results demonstrate the model’s effectiveness and reliability in real-world agricultural scenarios. We present an integrated framework that combines these ensemble techniques with Explainable AI (XAI). This technique ensures that the generated predictions are explainable, thus making it possible for stakeholders to understand and accept the recommendations offered.
35.
Garg, K.; Prabu, P.; Jayanthi, M. G.
In: Biomedical Materials and Devices, 2025, ISBN: 27314820 (ISSN); 27314812 (ISSN), (0).
@article{676,
title = {Efficient Breast Cancer Detection and Classification Using Rotation-Invariant Progressive Feedback Cosine Convolutional Neural Network with Tyrannosaurus Optimization Algorithm in Ultrasound Image},
author = {K. Garg and P. Prabu and M. G. Jayanthi},
url = {https://link.springer.com/article/10.1007/s44174-025-00539-3},
doi = {10.1007/s44174-025-00539-3},
isbn = {27314820 (ISSN); 27314812 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Biomedical Materials and Devices},
publisher = {Springer Nature},
abstract = {Breast cancer is the second most common cause of cancer-related deaths among women globally. Improved survival depends on a prompt and precise diagnosis, yet traditional ultrasound interpretation is time-consuming and error-prone. This study presents a new Rotation-Invariant Progressive Feedback Cosine Convolutional Neural Network (RiPFC-CNN-TOA) for the automatic and very effective diagnosis and classification of breast cancer in sonograms using the Tyrannosaurus Optimization Algorithm. First, UDIAT and BUSI dataset images are preprocessed with Fast Gradient Domain-Guided Image Filtering to improve contrast and eliminate noise while maintaining vital tumor features. Precise lesion segmentation is performed with a Machine learning-based Hybrid Mamba-Transformer model, capturing both spatial and sequential patterns of the images. For categorization, the RiPFC-CNN combines a Rotation-Invariant Attention Network and a Progressive Feedback Cosine CNN for feature-rich and orientation-invariant feature extraction. By modeling intelligent predator–prey behavior and fine-tuning weights, the Tyrannosaurus Optimization Algorithm also enhances the model's performance. The suggested approach obtains excellent performance: accuracy of 99.53% (BUSI) and 99.6% (UDIAT), precision over 99.3%, and large F1-scores and specificity, greatly enhancing diagnostic reliability with a significant reduction in false positives. This framework provides a robust, clinician-friendly algorithm for early breast cancer identification with excellent potential for real-world application in diagnostic pipelines.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Breast cancer is the second most common cause of cancer-related deaths among women globally. Improved survival depends on a prompt and precise diagnosis, yet traditional ultrasound interpretation is time-consuming and error-prone. This study presents a new Rotation-Invariant Progressive Feedback Cosine Convolutional Neural Network (RiPFC-CNN-TOA) for the automatic and very effective diagnosis and classification of breast cancer in sonograms using the Tyrannosaurus Optimization Algorithm. First, UDIAT and BUSI dataset images are preprocessed with Fast Gradient Domain-Guided Image Filtering to improve contrast and eliminate noise while maintaining vital tumor features. Precise lesion segmentation is performed with a Machine learning-based Hybrid Mamba-Transformer model, capturing both spatial and sequential patterns of the images. For categorization, the RiPFC-CNN combines a Rotation-Invariant Attention Network and a Progressive Feedback Cosine CNN for feature-rich and orientation-invariant feature extraction. By modeling intelligent predator–prey behavior and fine-tuning weights, the Tyrannosaurus Optimization Algorithm also enhances the model's performance. The suggested approach obtains excellent performance: accuracy of 99.53% (BUSI) and 99.6% (UDIAT), precision over 99.3%, and large F1-scores and specificity, greatly enhancing diagnostic reliability with a significant reduction in false positives. This framework provides a robust, clinician-friendly algorithm for early breast cancer identification with excellent potential for real-world application in diagnostic pipelines.
36.
Shirisha, M. S.; Prasad, M. S.; Rao, B. B.; Chetti, S.; Naikodi, C.
Integration of NIST-Aligned Secure IoT Data Management Framework Tinkercad, Flask, and ThingSpeak Proceedings
Institute of Electrical and Electronics Engineers Inc., 2024, ISBN: 9798350376517 (ISBN), (0).
@proceedings{374,
title = {Integration of NIST-Aligned Secure IoT Data Management Framework Tinkercad, Flask, and ThingSpeak},
author = {M. S. Shirisha and M. S. Prasad and B. B. Rao and S. Chetti and C. Naikodi},
doi = {10.1109/ICICEC62498.2024.10808927},
isbn = {9798350376517 (ISBN)},
year = {2024},
date = {2024-01-01},
journal = {1st International Conference on Innovations in Communications, Electrical and Computer Engineering, ICICEC 2024},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {This research paper presents a secure IoT data management framework that integrates Tinkercad, Flask, and ThingSpeak, while adhering to the NIST guidelines for data security. The Flask server plays a central role in implementing encryption, hashing, and decryption techniques to ensure the confidentiality, integrity, and authenticity of IoT data. By leveraging NIST-approved algorithms, such as AES for encryption and SHA-256 for hashing, and following secure key management practices, the framework provides a robust and standardized approach to the security of IoT data. The proposed solution demonstrates the effectiveness of using Flask for secure data processing in IoT environments while aligning with industry best practices. © 2024 IEEE.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
This research paper presents a secure IoT data management framework that integrates Tinkercad, Flask, and ThingSpeak, while adhering to the NIST guidelines for data security. The Flask server plays a central role in implementing encryption, hashing, and decryption techniques to ensure the confidentiality, integrity, and authenticity of IoT data. By leveraging NIST-approved algorithms, such as AES for encryption and SHA-256 for hashing, and following secure key management practices, the framework provides a robust and standardized approach to the security of IoT data. The proposed solution demonstrates the effectiveness of using Flask for secure data processing in IoT environments while aligning with industry best practices. © 2024 IEEE.
37.
Devi, K. K.; Kumar, J. P.
Sustainable Food Development Based on Ensemble Machine Learning Assisted Crop and Fertilizer Recommendation System Journal Article
In: Journal of Machine and Computing, vol. 4, pp. 317-326,, 2024, ISBN: 27891801 (ISSN), (2).
@article{11,
title = {Sustainable Food Development Based on Ensemble Machine Learning Assisted Crop and Fertilizer Recommendation System},
author = {K. K. Devi and J. P. Kumar},
doi = {10.53759/7669/jmc202404030},
isbn = {27891801 (ISSN)},
year = {2024},
date = {2024-01-01},
journal = {Journal of Machine and Computing},
volume = {4},
pages = {317-326,},
publisher = {AnaPub Publications},
abstract = {Agriculture is the most vital sector for the global food supply, and it also provides raw materials for other types of industries. A crop recommendation system is essential for farmers who want to get the most out of their crop-choosing decisions. Over the last several decades, the world's ability to produce food has grown substantially owing to the extensive usage of fertilizers. Therefore, there has to be a more eco-friendly and effective way to utilize fertilizers that include nitrogen (N), phosphorous (P), and potassium (K) to ensure food security. For the reason, this study proposes an ensemble machine learning-assisted crop and fertilizer recommendation system (EML-CFRS) to maximize agricultural output while ensuring the correct use of mineral resources. The research used a dataset obtained from the Kaggle repository like that people can assess several distinct ML algorithms. The databases include data on three climate variables-temperature, rainfall, and humidity-and information on NPK and soil pH. The yields agricultural crops were used to train these models, including Decision Tree, KNN, XGBoost, Support Vector Machine, and Random Forest. Depending on the current weather and soil conditions, the trained model may then recommend the optimal fertiliser for a certain crop. Predicting the ideal kind and quantity of fertilizer for different crops was accomplished with a 96.5% accuracy rate by our suggested strategy. © 2024 The Authors.},
note = {2},
keywords = {CSE},
pubstate = {published},
tppubtype = {article}
}
Agriculture is the most vital sector for the global food supply, and it also provides raw materials for other types of industries. A crop recommendation system is essential for farmers who want to get the most out of their crop-choosing decisions. Over the last several decades, the world's ability to produce food has grown substantially owing to the extensive usage of fertilizers. Therefore, there has to be a more eco-friendly and effective way to utilize fertilizers that include nitrogen (N), phosphorous (P), and potassium (K) to ensure food security. For the reason, this study proposes an ensemble machine learning-assisted crop and fertilizer recommendation system (EML-CFRS) to maximize agricultural output while ensuring the correct use of mineral resources. The research used a dataset obtained from the Kaggle repository like that people can assess several distinct ML algorithms. The databases include data on three climate variables-temperature, rainfall, and humidity-and information on NPK and soil pH. The yields agricultural crops were used to train these models, including Decision Tree, KNN, XGBoost, Support Vector Machine, and Random Forest. Depending on the current weather and soil conditions, the trained model may then recommend the optimal fertiliser for a certain crop. Predicting the ideal kind and quantity of fertilizer for different crops was accomplished with a 96.5% accuracy rate by our suggested strategy. © 2024 The Authors.
38.
Farooq, U.; Reddy, K. K. S.; Shishira, K. S.; Jayanthi, M. G.; Kannadaguli, P.
Comparing Hindustani Music Raga Prediction Systems using DL and ML Models Proceedings
Institute of Electrical and Electronics Engineers Inc., 2024, ISBN: 979-835037250-2 (ISBN), (1).
@proceedings{9,
title = {Comparing Hindustani Music Raga Prediction Systems using DL and ML Models},
author = {U. Farooq and K. K. S. Reddy and K. S. Shishira and M. G. Jayanthi and P. Kannadaguli},
doi = {10.1109/ICETCS61022.2024.10543647},
isbn = {979-835037250-2 (ISBN)},
year = {2024},
date = {2024-01-01},
journal = {International Conference on Emerging Technologies in Computer Science for Interdisciplinary Applications, ICETCS 2024},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {This project aims to employ DL and ML techniques to predict ragas in Indian classical music accurately. The focus is on developing a system that can process audio recordings and make precise raga predictions. The classification task utilizes CNN and RNN networks, to enhance performance. Extensive evaluation using diverse recordings is conducted, comparing the framework against traditional methods. The outcomes of this project has potential applications for music analysis, archiving, recommendation systems, and education in Indian classical music. The developed raga prediction framework can serve as a valuable tool for automatic raga identification. Additionally, it contributes to the field of music information retrieval by showcasing the capabilities of DL/ML techniques in tackling musical tasks. © 2024 IEEE.},
note = {1},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
This project aims to employ DL and ML techniques to predict ragas in Indian classical music accurately. The focus is on developing a system that can process audio recordings and make precise raga predictions. The classification task utilizes CNN and RNN networks, to enhance performance. Extensive evaluation using diverse recordings is conducted, comparing the framework against traditional methods. The outcomes of this project has potential applications for music analysis, archiving, recommendation systems, and education in Indian classical music. The developed raga prediction framework can serve as a valuable tool for automatic raga identification. Additionally, it contributes to the field of music information retrieval by showcasing the capabilities of DL/ML techniques in tackling musical tasks. © 2024 IEEE.
39.
Sanath, G.; Deekshitha, S.; Shravya, H.; Jayanthi, M. G.; Kannadaguli, P.
Aquify:AI-Enhanced Predictive Analytics Toolkit For Water Contamination Proceedings
Institute of Electrical and Electronics Engineers Inc., 2024, ISBN: 979-833152853-9 (ISBN), (0).
@proceedings{351,
title = {Aquify:AI-Enhanced Predictive Analytics Toolkit For Water Contamination},
author = {G. Sanath and S. Deekshitha and H. Shravya and M. G. Jayanthi and P. Kannadaguli},
url = {https://ieeexplore.ieee.org/document/10748475},
doi = {10.1109/I4C62240.2024.10748475},
isbn = {979-833152853-9 (ISBN)},
year = {2024},
date = {2024-01-01},
journal = {5th International Conference on Circuits, Control, Communication and Computing, I4C 2024},
pages = {177-182,},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Effective monitoring of water quality remains a problem, in real-time contamination detection and prediction. Despite advancements in the sector, many existing approaches continue to rely on labor-intensive laboratory testing, which limits their capacity to give rapid and accurate results. By leveraging advanced ML/DL models, it is possible to build a more robust and adaptable platform for real-time analysis, prediction, and decision support. Approaches based on models such as Random Forest, XGBoost, CNN, RNN, LSTM, and GRU have been good with the accuracy of water quality predictions. They help identify pollution sources and diagnose water quality issues, providing essential information to stakeholders in sustainable water resource management. The ML and DL approaches help improve water quality monitoring and boost future research in environmental sustainability and natural resource management.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Effective monitoring of water quality remains a problem, in real-time contamination detection and prediction. Despite advancements in the sector, many existing approaches continue to rely on labor-intensive laboratory testing, which limits their capacity to give rapid and accurate results. By leveraging advanced ML/DL models, it is possible to build a more robust and adaptable platform for real-time analysis, prediction, and decision support. Approaches based on models such as Random Forest, XGBoost, CNN, RNN, LSTM, and GRU have been good with the accuracy of water quality predictions. They help identify pollution sources and diagnose water quality issues, providing essential information to stakeholders in sustainable water resource management. The ML and DL approaches help improve water quality monitoring and boost future research in environmental sustainability and natural resource management.
40.
Shaheema, S. B.; Ns, B.; Jose, P.; Albert, I. E.; Anorelin, A.; Tony, E. I.
AI-Powered Traffic Surveillance: License Plate Recognition with Non-Helmet Detection Using YOLOv8 Proceedings
Institute of Electrical and Electronics Engineers Inc., 2024, ISBN: 9798350376135 (ISBN), (0).
@proceedings{370,
title = {AI-Powered Traffic Surveillance: License Plate Recognition with Non-Helmet Detection Using YOLOv8},
author = {S. B. Shaheema and B. Ns and P. Jose and I. E. Albert and A. Anorelin and E. I. Tony},
doi = {10.1109/SPICES62143.2024.10779944},
isbn = {9798350376135 (ISBN)},
year = {2024},
date = {2024-01-01},
journal = {2024 IEEE International Conference on Signal Processing, Informatics, Communication and Energy Systems, SPICES 2024},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Helmet detection(HD) is critical for modern intelligent systems to locate and identify various bike riders without helmets. HD is challenging due to several issues, such as the helmet's shapes, color, and designs, uneven outlines, angle changes, and occlusion. Automatic license plate recognition (ALPR) technology recognizes the optical character on the number plate and locates the bike riders without a helmet by detecting the license plate. This research focuses on identifying the helmet detection of the bike images captured by CCTV in India. This paper presents an HD approach to address the problems as mentioned earlier. The suggested strategy consists of the following steps. First, the You Only Look Once version 8 (YOLOv8) network detects an image of a bike that shows up in an input image. The helmet is next detected within the indicated bike using morphological techniques, and finally, the network is employed to recognize license plates. The simulation results test shows an accuracy rate of 99.34% and character recognition of 98.76% , respectively. © 2024 IEEE.},
note = {0},
keywords = {CSE},
pubstate = {published},
tppubtype = {proceedings}
}
Helmet detection(HD) is critical for modern intelligent systems to locate and identify various bike riders without helmets. HD is challenging due to several issues, such as the helmet's shapes, color, and designs, uneven outlines, angle changes, and occlusion. Automatic license plate recognition (ALPR) technology recognizes the optical character on the number plate and locates the bike riders without a helmet by detecting the license plate. This research focuses on identifying the helmet detection of the bike images captured by CCTV in India. This paper presents an HD approach to address the problems as mentioned earlier. The suggested strategy consists of the following steps. First, the You Only Look Once version 8 (YOLOv8) network detects an image of a bike that shows up in an input image. The helmet is next detected within the indicated bike using morphological techniques, and finally, the network is employed to recognize license plates. The simulation results test shows an accuracy rate of 99.34% and character recognition of 98.76% , respectively. © 2024 IEEE.
