K.SELVA SHEELA

@sreesakthi.edu.in

Assistant Professor in CSE DEPT
SREE SAKTHI ENGINEERING COLLEGE

K.SELVA SHEELA


                     Download Compact PDF  
https://researchid.co/selvasheela
SHEELA MCA,ME, Phd, act as an Head of the department / Associate Professor of Computer Science and Engineering department of Sree Sakthi Engineering College , Karamadai She is having more than 14 years of teaching experience in academic and research Her research interest include Bigdata ,Opinion Mining , Artifical Intelligence and Distuributed Computing. She has 7 international Publications and present more than 25 papers international conferences

EDUCATION

MCA , ME , (PHD)

RESEARCH INTERESTS

BIGDATA , IOT, NETWORK SECURITY , CLOUD COMPUTING
12

Scopus Publications

59

Scholar Citations

4

Scholar h-index

3

Scholar i10-index

Scopus Publications

  • GT-SOH: A Graph-Transformer Contrastive Learning Framework with Starfish Optimization for Accurate State-of-Health Battery Prediction in EV
    R. Natarajan, Jeya Prakash Kadambarajan, K. Selva Sheela
    Ssrg International Journal of Electronics and Communication Engineering, 2025
    In specific, accurate, and precise prediction of the State-of-Health (SOH) of the Electric Vehicle (EV) battery system will be crucial to ensure battery safety, longevity, and optimal management of energy. Existing SoH estimation models, which include Machine Learning (ML) and Deep Learning (DL) schemes, most often struggle in capturing the most complex temporal and spatial dependencies in the degradation of the battery. In this paper, a novel Graph-Transformer Contrastive Learning (GT-SoH) framework, which incorporates Graph Neural Networks (GNNs) termed Transformer-based temporal modeling, Contrastive self-learning, and Starfish Optimization Algorithm (SFOA) for hyperparameter tuning, is proposed and is denoted as the (GT-SOH-SFOA) framework. A GNN model is responsible for capturing spatial interdependencies among battery cells, whereas a Transformer encoder models GNN patterns. A contrastive learning function is used for enhancing the generalizability of learning a robust representation of features from unlabeled battery datasets. In addition, SFOA is employed to tune the hyperparameters, thus ensuring optimal performance for balancing exploitation and exploration in the process of optimization. The hybrid loss function, which integrates Mean Absolute Error (MAE) loss and contrastive loss, ensures precise SOH estimation, thus reducing overfitting. An experimental evaluation is carried out for various metrics like Mean Absolute Error (MAE), R2, RMSE, and Max Error on four datasets, like Musoshi, NASA, Stanford, and the BMW i3 battery dataset, and outcomes attained demonstrate that the GT-SOH-SFOA proposed model outperforms existing models compared, thereby offering high prediction accuracy and robustness. Therefore, it is concluded that the proposed scheme offers a scalable, interpretable, and optimized solution for real – time battery health monitoring in EVs.
  • Intelligent soil fertility forecasting using enhanced STGNN and hybrid swarm-based optimization
    C P Thamil Selvi, M. Manimaraboopathy, M. Jeyalakshmi, G Narmadha, Selva Sheela K
    Results in Engineering, 2025
    Soil fertility plays a key role in sustainable agricultural productivity and environmental health. Existing soil assessment techniques are labour-intensive and time-consuming. Also, they are limited in their spatial and temporal coverage. To address these challenges, in this paper, a deep learning-based Enhanced Spatiotemporal Graph Neural Network (E-STGNN) is proposed. The model captures spatial dependencies through graph convolutional layers and temporal dynamics using recurrent neural structures. In addition, the E-STGNN model is optimized using a hybrid Particle Swarm Optimization-guided Red Kite Optimization (PSO-RKO) algorithm to improve model performance. Experimental results on Kaggle dataset shows that the proposed model achieves an exceptional accuracy of 98.9%, with corresponding improvements in precision (98.54%), recall (98.68%), and F1-score (98.60%).
  • Shifted split-merge segmentation and fuzzy-guided generative adversarial network underwater object detection
    K Selva Sheela, S Vinoth Kumar, Saman M Almufti, R Lakshmana Kumar
    Intelligent Data Analysis, 2025
    Marine object localization from UAVs is required for different areas, including marine research, environment monitoring and topographic surveys. Herein, we introduce a new undersea object detection method; synthesizing shifted split-merge segmentation and fuzzy-oriented generative adversarial networks. Split-merge segmentation is the region-based model, where this process is achieved via splitting and merging regions based on specified similarity measures. The previous split-merge segmentation algorithm was modified to employ a shifted window approach that is better at detecting undersea objects with changing shapes and sizes to address this issue. In addition to Guiding the segmentation quality of underwater object detection, a Fuzzy-Guided Generative Adversarial Network (FG-GAN) is proposed. The generator network aimed to produce artificially photographed images beneath the water, and the discriminator network was used to differentiate between real and fabricated pictures. The generator system is trained to use a fuzzy loss function with fuzzy membership functions to explain the level of uncertainty and vagueness in the underwater environment by controlling the behaviour of underwater entities. We positioned the above-described method and compared it to the traditionally used image partition and object detection approaches. The outcomes from these experiments indicate that our proposed method is more accurate than the existing approaches in segmenting the objects and identifying the objects accurately with 95% and has a reduced loss of 0.3. The proposed approach could be applied in a broad spectrum of underwater facilities such as marine hydrology, work with remote sensing equipment and underwater robotics.
  • Multi-Sensor Data Fusion with Explainable AI for Anomaly Detection in Industrial CPS
    K Selva Sheela, Gokila Deepa G, Pavithra A, S Jeevanandham, M Sundarrajan, Mani Deepak Choudhry
    Proceedings of the 2025 1st International Conference on Advances in Engineering and Computing Technologies for Sustainable Development Aectsd 2025, 2025
    With the emergence of the Cyber-Physical Systems (CPS) and the Industrial Internet of Things (IIoT) being connected to the systems, the issues of reliability, safety, and performance are urgent. The data produced by sensors is either enormous or too dominant, to the extent that anomalies in such environments are further complicated and more difficult to detect. The classical methods of anomaly detection, including rule-based diagnostic techniques, threshold-based methods, and classical machine learning models such as Support Vector Machines (SVMs) and Random Forests, cannot detect complex temporal patterns and interactions. Lastly, these models cannot be deployed in a mission-critical setting, as agile or explainable, because operational trust and regulatory compliance require explainable decisions. To address these limitations, this study provided a new framework for anomaly detection that relies on the fusion of multiple sensors and a hybrid deep learning model consisting of Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU), and refines it with explainable AI, including SHAP. Both simulated and semi-real industrial data, under different operating conditions and failure scenarios, are experimentally verified using the semi-real and simulated data. The proposed CNN-GRU hybrid model was found to be accurate (94), has a precision of (92), a recall of (95) and an F1-score of (0.935). Lastly, these findings demonstrate a considerable improvement over the state-of-the-art models, i. e., SVM (accuracy 85%), Random Forest (accuracy 87%) and LSTM (accuracy 91%), not only in terms of the capacity of their models to generalize, be sensitive and interpretable, but also in terms of the capacity of the proposed framework to do better than such results.
  • Enhancing liver tumor segmentation with UNet-ResNet: Leveraging ResNet's power
    K. Selva Sheela, Vivek Justus, Renas Rajab Asaad, R. Lakshmana Kumar
    Technology and Health Care, 2024
    BACKGROUND: Liver cancer poses a significant health challenge due to its high incidence rates and complexities in detection and treatment. Accurate segmentation of liver tumors using medical imaging plays a crucial role in early diagnosis and treatment planning. OBJECTIVE: This study proposes a novel approach combining U-Net and ResNet architectures with the Adam optimizer and sigmoid activation function. The method leverages ResNet’s deep residual learning to address training issues in deep neural networks. At the same time, U-Net’s structure facilitates capturing local and global contextual information essential for precise tumor characterization. The model aims to enhance segmentation accuracy by effectively capturing intricate tumor features and contextual details by integrating these architectures. The Adam optimizer expedites model convergence by dynamically adjusting the learning rate based on gradient statistics during training. METHODS: To validate the effectiveness of the proposed approach, segmentation experiments are conducted on a diverse dataset comprising 130 CT scans of liver cancers. Furthermore, a state-of-the-art fusion strategy is introduced, combining the robust feature learning capabilities of the UNet-ResNet classifier with Snake-based Level Set Segmentation. RESULTS: Experimental results demonstrate impressive performance metrics, including an accuracy of 0.98 and a minimal loss of 0.10, underscoring the efficacy of the proposed methodology in liver cancer segmentation. CONCLUSION: This fusion approach effectively delineates complex and diffuse tumor shapes, significantly reducing errors.
  • Sentiment Analysis of Text and Emoji using Machine Learning Algorithms
    Rubini P. E, K. Selva Sheela, R. Delshi Howsalya Devi, R. Yamuna, K. Penyameen, S. Gowdhamkumar
    Tqcebt 2024 2nd IEEE International Conference on Trends in Quantum Computing and Emerging Business Technologies 2024, 2024
    A major form of communication after the speech is text. Unlike direct communication through a speech where the tone can be analyzed depending on the pitch and frequency, text analysis is based on the frequency of words in a particular text. A recent survey conducted shows that around 80% of business data is in textual format. All industry and business users including call centers, online reviews, customer surveys, etc., work with raw textual data. This untapped data when mined and analyzed can provide great insights to the organization. Consumers are interested in other’s opinions and experiences. Statistics reveal that at least 90% of people are influenced by what they read and hence the negative comments take a huge toll. Over the recent years, sites have started collecting customer reviews for their products. Companies use ‘Text Analysis’ to manage their content. Textual data when converted into data slices that are easy to automate, processes like product development, decision making, marketing optimization, and business intelligence take place easily. When the text analysis process is automated, it helps analyze large chunks of data within a short period of time and with lesser resources when compared to manual analysis. Automated text analysis is also proven to be faster and scalable along with consistent results. The purpose of this work is to develop a text analysis algorithm that will classify text into either one of the three categories: Positive, Negative, or Neutral. Once the received text is cleaned and pre-processed, it can be classified into one of the categories based on classification techniques like Naive Bayes, Support Vector Machine, etc., although it is observed that the Naive Bayes approach produces better accuracy. Similar to text, emoji can also be converted to its respective textual sequence and then classified.
  • Reinforcement Learning in Cognitive Robots for Autonomous Path Planning
    Archana Das, R. Indu Poornima, G. Gokila Deepa, K. Selva Sheela, Akshya J, Mani Deepak Choudhry
    Proceedings of the 5th International Conference on Data Intelligence and Cognitive Informatics Icdici 2024, 2024
    Cognitive robots are intelligent systems that learn from their environment, adapt to dynamic changes, and make decisions without a human's direct intervention. This research studies the integration of Reinforcement Learning in cognitive robots with Q-Learning for autonomous path planning. The study addresses the most critical issues in traditional A* and Dynamic Programming algorithms by devising path-planning techniques, mainly the lack of adaptability and computational efficiency in real-time and unpredictable environments. Our approach is based on Q-Learning, which is a model-free RL algorithm allowing the robots to find paths autonomously by optimizing their paths and avoiding obstacles. The Q-Learning algorithm provides the possibility of learning optimal policies for the robot through iterative interaction with its environment balanced between exploration and exploitation of the decision process. A reward system is utilized by the proposed model, encouraging the robot to explore shorter, collision-free paths and adjust based on feedback in real time from its surroundings. The model was found to be significantly better than its counterparts. In this context, as indicated, the Q-Learning model runs at 11.5 seconds, faster than A* at 18.3 seconds and Dynamic Programming at 16.7 seconds, yet with an accuracy of 94% and was found to have the highest collision avoidance rate at 98%. Additionally, the adaptability of the model about the environment presents a marked difference in terms of path length optimization, having done so with a mean path length of 12.3, compared to approaches or models. The robustness and scalability of the Q-Learning model make it highly applicable in real-world applications.
  • Early fire detection technique for human being using deep learning algorithm
    Kannan Deeba, Sattianadan Dasarathan, Srinivasa Rao Kandula, Krishnasamy Selva Sheela, Ravindran Ramkumar, Nagarajan Ashokkumar, Dhandapani Karthikeyan
    Indonesian Journal of Electrical Engineering and Computer Science, 2023
    Fire and smoke detection in today’s world is a must, especially in clustered areas where a quick response can prevent significant damages and save lives. Early detection plays a significant role in preventing the fire from spreading by alerting the emergency response personnel. It may not be possible to install traditional fire and smoke detectors everywhere. As a result, incorporating fire and smoke detection into existing closed circuit television (CCTV) systems in various places can provide a warning to the appropriate authorities, allowing for quick action to prevent the fire from spreading. This work aims in developing an early fire and smoke prediction model with CCTV footage images and video frames. The images and videos are collected from multiple datasets available online. A convolutional neural network (CNN) model is developed for early detection and prevention of the spreading of fire and compares it with transfer learning models ResNet50 and VGG19. The model obtain an accuracy of around 94% using CNN model, 95% using VGG19 and 98% using ResNet 50. A model with high accuracy can replace traditional fire detection systems which can be both cost-effective and easy to implement to existing surveillance cameras.
  • Effective Customer Review Analysis Using Combined Capsule Networks with Matrix Factorization Filtering
    K. Selvasheela, A. M. Abirami, Abdul Khader Askarunisa
    Computer Systems Science and Engineering, 2023
    Nowadays, commercial transactions and customer reviews are part of human life and various business applications. The technologies create a great impact on online user reviews and activities, affecting the business process. Customer reviews and ratings are more helpful to the new customer to purchase the product, but the fake reviews completely affect the business. The traditional systems consume maximum time and create complexity while analyzing a large volume of customer information. Therefore, in this work optimized recommendation system is developed for analyzing customer reviews with minimum complexity. Here, Amazon Product Kaggle dataset information is utilized for investigating the customer review. The collected information is analyzed and processed by batch normalized capsule networks (NCN). The network explores the user reviews according to product details, time, price purchasing factors, etc., ensuring product quality and ratings. Then effective recommendation system is developed using a butterfly optimized matrix factorization filtering approach. Then the system’s efficiency is evaluated using the Rand Index, Dunn index, accuracy, and error rate.
  • Twitter Media Sentiment Analysis to Convert Non-Informative to Informative Using QER
    C. P. Thamil Selvi, P. Muneeshwari, K. Selvasheela, D. Prasanna
    Intelligent Automation and Soft Computing, 2023
    The term sentiment analysis deals with sentiment classification based on the review made by the user in a social network. The sentiment classification accuracy is evaluated using various selection methods, especially those that deal with algorithm selection. In this work, every sentiment received through user expressions is ranked in order to categorise sentiments as informative and non-informative. In order to do so, the work focus on Query Expansion Ranking (QER) algorithm that takes user text as input and process for sentiment analysis and finally produces the results as informative or non-informative. The challenge is to convert non-informative into informative using the concepts of classifiers like Bayes multinomial, entropy modelling along with the traditional sentimental analysis algorithm like Support Vector Machine (SVM) and decision trees. The work also addresses simulated annealing along with QER to classify data based on sentiment analysis. As the input volume is very fast, the work also addresses the concept of big data for information retrieval and processing. The result comparison shows that the QER algorithm proved to be versatile when compared with the result of SVM. This work uses Twitter user comments for evaluating sentiment analysis.
  • Creating the effective customer review analysis system using the batch normalization with abc collaborative recommendation system
    Selva Sheela Krishnasamy, Ariyur Mahadevan Abirami, Abdulkhader Askarunisa
    Comptes Rendus De L Academie Bulgare Des Sciences, 2021
  • A survey of financial forecasting and customer analysis in banking institution
    International Journal of Applied Engineering Research, 2015

RECENT SCHOLAR PUBLICATIONS

  • Multi-Sensor Data Fusion with Explainable AI for Anomaly Detection in Industrial CPS
    KS Sheela, A Pavithra, S Jeevanandham, M Sundarrajan, MD Choudhry
    2025 First International Conference of Advances in Engineering and Computing … , 2025
    2025.0
  • Shifted split-merge segmentation and fuzzy-guided generative adversarial network underwater object detection
    KS Sheela, SV Kumar, SM Almufti, RL Kumar
    Intelligent Data Analysis 29 (4), 1037-1061 , 2025
    2025.0
    Citations: 12
  • Enhancing liver tumor segmentation with UNet-ResNet: Leveraging ResNet’s power
    KS Sheela, V Justus, RR Asaad, RL Kumar
    Technology and Health Care 33 (1), 1-15 , 2025
    2025.0
    Citations: 16
  • Reinforcement Learning in Cognitive Robots for Autonomous Path Planning
    A Das, RI Poornima, GG Deepa, KS Sheela, MD Choudhry
    2024 5th International Conference on Data Intelligence and Cognitive … , 2024
    2024.0
  • Sentiment Analysis of Text and Emoji using Machine Learning Algorithms
    KS Sheela, RDH Devi, R Yamuna, K Penyameen, S Gowdhamkumar
    2024 International Conference on Trends in Quantum Computing and Emerging … , 2024
    2024.0
    Citations: 1
  • TEMPORAL GAN ENSEMBLE WITH BAGGING FOR ROBUST INFORMATION SECURITY IN IOT SENSOR NETWORKS
    RR S. Roselin Mary , K. Selva Sheela, Srinivasa Rao Kandula
    ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY 14 (03), 3024-28 , 2023
    2023.0
  • Early fire detection technique for human being using deep learning algorithm
    DK Kannan Deeba , Sattianadan Dasarathan , Srinivasa Rao Kandula ...
    Indonesian Journal of Electrical Engineering and Computer Science 31 (3 … , 2023
    2023.0
    Citations: 1
  • Twitter Media Sentiment Analysis to Convert Non-Informative to Informative Using QER.
    CP Selvi, P Muneeshwari, K Selvasheela, D Prasanna
    Intelligent Automation & Soft Computing 35 (3) , 2023
    2023.0
    Citations: 24
  • Effective Customer Review Analysis Using Combined Capsule Networks with Matrix Factorization Filtering.
    K Selvasheela, AM Abirami, AK Askarunisa
    Computer Systems Science & Engineering 44 (3) , 2023
    2023.0
    Citations: 5
  • Natural Language Processing for Sentiment Analysis in Social Media
    HP C.P.Thamil Selvi , K.Selva sheela
    2023.0
  • Creating the Effective Customer Review Analysis System Using the Batch Normalization with ABC Collaborative Recommendation System
    SS Krishnasamy, AM Abirami, A Askarunisa
    COMPTES RENDUS DE L ACADEMIE BULGARE DES SCIENCES 74 (5), 756-766 , 2021
    2021.0
  • Honey Encryption Based Data Store in Cloud
    D Rekha, KS Sheela
    2017.0
  • GT-SOH: A Graph-Transformer Contrastive Learning Framework with Starfish Optimization for Accurate State-of-Health Battery Prediction in EV
  • Intelligent soil fertility forecasting using enhanced STGNN and hybrid swarm-based optimization

MOST CITED SCHOLAR PUBLICATIONS

  • Twitter Media Sentiment Analysis to Convert Non-Informative to Informative Using QER.
    CP Selvi, P Muneeshwari, K Selvasheela, D Prasanna
    Intelligent Automation & Soft Computing 35 (3) , 2023
    2023.0
    Citations: 24
  • Enhancing liver tumor segmentation with UNet-ResNet: Leveraging ResNet’s power
    KS Sheela, V Justus, RR Asaad, RL Kumar
    Technology and Health Care 33 (1), 1-15 , 2025
    2025.0
    Citations: 16
  • Shifted split-merge segmentation and fuzzy-guided generative adversarial network underwater object detection
    KS Sheela, SV Kumar, SM Almufti, RL Kumar
    Intelligent Data Analysis 29 (4), 1037-1061 , 2025
    2025.0
    Citations: 12
  • Effective Customer Review Analysis Using Combined Capsule Networks with Matrix Factorization Filtering.
    K Selvasheela, AM Abirami, AK Askarunisa
    Computer Systems Science & Engineering 44 (3) , 2023
    2023.0
    Citations: 5
  • Sentiment Analysis of Text and Emoji using Machine Learning Algorithms
    KS Sheela, RDH Devi, R Yamuna, K Penyameen, S Gowdhamkumar
    2024 International Conference on Trends in Quantum Computing and Emerging … , 2024
    2024.0
    Citations: 1
  • Early fire detection technique for human being using deep learning algorithm
    DK Kannan Deeba , Sattianadan Dasarathan , Srinivasa Rao Kandula ...
    Indonesian Journal of Electrical Engineering and Computer Science 31 (3 … , 2023
    2023.0
    Citations: 1
  • Multi-Sensor Data Fusion with Explainable AI for Anomaly Detection in Industrial CPS
    KS Sheela, A Pavithra, S Jeevanandham, M Sundarrajan, MD Choudhry
    2025 First International Conference of Advances in Engineering and Computing … , 2025
    2025.0
  • Reinforcement Learning in Cognitive Robots for Autonomous Path Planning
    A Das, RI Poornima, GG Deepa, KS Sheela, MD Choudhry
    2024 5th International Conference on Data Intelligence and Cognitive … , 2024
    2024.0
  • TEMPORAL GAN ENSEMBLE WITH BAGGING FOR ROBUST INFORMATION SECURITY IN IOT SENSOR NETWORKS
    RR S. Roselin Mary , K. Selva Sheela, Srinivasa Rao Kandula
    ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY 14 (03), 3024-28 , 2023
    2023.0
  • Natural Language Processing for Sentiment Analysis in Social Media
    HP C.P.Thamil Selvi , K.Selva sheela
    2023.0
  • Creating the Effective Customer Review Analysis System Using the Batch Normalization with ABC Collaborative Recommendation System
    SS Krishnasamy, AM Abirami, A Askarunisa
    COMPTES RENDUS DE L ACADEMIE BULGARE DES SCIENCES 74 (5), 756-766 , 2021
    2021.0
  • Honey Encryption Based Data Store in Cloud
    D Rekha, KS Sheela
    2017.0
  • GT-SOH: A Graph-Transformer Contrastive Learning Framework with Starfish Optimization for Accurate State-of-Health Battery Prediction in EV
  • Intelligent soil fertility forecasting using enhanced STGNN and hybrid swarm-based optimization