BIJAYA KUMAR SETHI

RESEARCH, TEACHING, or OTHER INTERESTS

Artificial Intelligence, Artificial Intelligence, Computer Science Applications

Scopus Publications

Mitigating Threats in PHY-Layer Authentication: A Proactive Defense Against Membership Inference Attacks in Wireless Signal Classifiers
D. Madhuri, V. Nikitha Reddy, M. Keerthi Reddy, V. N. L. N. Murthy, Saroja Kumar Rout, and Bijaya Kumar Sethi
Springer Nature Switzerland



A Precision Tessellated Fundus Detection: Leveraging Color and Texture Features with SVM Classification
Kachi Anvesh, Bharati M. Reshmi, Saroj Kumar Rout, Bijaya Kumar Sethi, and Satya Sobhan Panigrahi
Springer Nature Singapore



Prostate Cancer Prediction Using Convolutional Neural Networks
Bijaya Kumar Sethi, Debabrata Singh, and Saroja Kumar Rout
IEEE
Cancer-related mortality in men is highest among men who suffer from prostate cancer. The lack of clarity and consistency of early symptoms often makes diagnosis a challenge in the later stages (stages III and IV). Existing diagnostic techniques face challenges such as subjectivity, variability between observers, and lengthy testing processes involving biomarkers, biopsies, and imaging tests. This paper introduces a novel convolutional neural network (CNN) algorithm for prostate cancer diagnosis and prediction to overcome these drawbacks. An additional dataset of histopathology images was used to train and validate the system before it was put through its learning phase. In the study, 95.12% of cancerously derived cells were identified correctly and 93.02% were identified correctly, a remarkable accuracy of 98.07%. As a result of this study, Various challenges associated with expert evaluations by humans were successfully addressed, including higher misclassification rates, interdependencies between observers, and lengthy analysis periods. Prostate cancer diagnosis and prognosis have been made much simpler and faster by this research. Moving forward, to optimize the effectiveness of our proposed method, future investigations should explore the latest developments and innovations in this field.


Diabetes Prediction using Hybrid Machine Learning Techniques
Bhukya Devendar Naik, Anumula Spoorthy, Sakshi Baherji, Saroja Kumar Rout, and Bijaya Kumar Sethi
IEEE
The rise in blood glucose levels is the primary factor contributing to the development of diabetes. Given the significance of preventing diabetes or delaying its onset, despite numerous efforts utilizing machine learning for medical diagnostics, there remains a notable gap in research concerning long-term disease prediction, especially for type 2 diabetes. However, the traditional method of diagnosing diabetes involves patients undergoing blood glucose tests administered by doctors, which can be limited by clinical resources. Many patients consequently encounter delays in getting a diagnosis. To create a predictive model for diabetes diagnosis, this study used six traditional machine learning models: boosting, neural networks, decision trees, random forests, logistic regression, and support vector machines. The study employed machine learning (ML) algorithms to predict diabetes using an authentic dataset from Safety Pressure Primary Health Care. With a validation accuracy of 84%, the study offers important new information on who is most likely to develop type 2 diabetes. By precisely predicting the type of diabetes and examining the importance of each indication in the prediction process, the goal of this study is to improve the accuracy of diabetes prediction.


Automated Interview Evaluation System Using RoBERTa Technology
G. Sri Harsh, Y. Sai Sri Vivek, Maneesha. P, Saroja Kumar Rout, S. Ranjith Reddy, and Bijaya Kumar Sethi
IEEE
This research paper delves into the realm of automated interview evaluation, employing state-of-the-art natural language processing models. The primary focus lies in comparing the effectiveness of TF-IDF, BERT, RoBERTa, Jaccard similarity, and Word2Vec models in assessing interview responses. The paper commences by providing a comprehensive definition of automated interview evaluation, highlighting the significance of efficient and unbiased candidate assessment. The proposed methodology involves the utilization of RoBERTa, a robust transformer-based model, to analyze and score interview responses. Through meticulous experimentation and evaluation, the research scrutinizes the performance of each model, examining its ability to capture contextual nuances and semantic understanding. The final results reveal the superior efficacy of RoBERTa over the other models, demonstrating its proficiency in evaluating interview responses and emphasizing its potential for enhancing automated hiring processes. This study contributes valuable insights into the evolving field of natural language processing and automated recruitment systems.


Breast Cancer Detection Using Convolutional Neural Network
Sai Sudharshan Saniganti, Shriyans Reddy Gaddam, Srinivas Reddy Eppa, Saroja Kumar Rout, Bijaya Kumar Sethi, and Bhaskerreddy Kethireddy
IEEE
This research endeavors by combining transfer learning techniques with Convolutional Neural Networks (CNNs), this study aims to improve the categorization of breast cancer. The primary objective is to improve diagnostic precision in detecting malignancies from mammographic images, ultimately impacting clinical decision-making and patient care positively. The study employs a robust methodology utilizing a diverse dataset for training and validation. Transfer learning optimizes CNNs' efficiency, fine-tuning the architecture to adapt to breast cancer detection nuances. Rigorous training-validation cycles refine the model, ensuring generalizability across diverse datasets. The automated system minimizes subjective variability, contributing to a more objective diagnostic process. Scalability is achieved by designing the model to handle large volumes of mammographic images, a critical feature for widespread implementation. The integration of CNNs and transfer learning yields promising results, demonstrating a substantial improvement in accuracy compared to existing methods. Automation significantly reduces diagnosis time, while introduced objectivity minimizes result variability. The property of scalability shows promise for broad use since it works well in managing massive amounts of images. These outcomes highlight the viability and effectiveness of the suggested strategy in improving the diagnosis of breast cancer. In conclusion, the developed model, combining CNNs and transfer learning, represents a significant advancement with the potential to revolutionize clinical decision-making and patient care, offering a more accurate, efficient, and widely applicable approach to breast cancer diagnosis.


Long Short-Term Memory-Deep Belief Network-Based Gene Expression Data Analysis for Prostate Cancer Detection and Classification
Bijaya Kumar Sethi, Debabrata Singh, Saroja Kumar Rout, and Sandeep Kumar Panda
Institute of Electrical and Electronics Engineers (IEEE)
Prostate cancer (PRC) is the major reason of mortality globally. Early recognition and classification of PRC become essential to enhance the quality of healthcare services. A newly established deep learning (DL) and machine learning (ML) approach with different optimization tools can be employed to classify accurately of PRC accurately using microarray gene expression data (GED). Though the microarray data structures are important to diagnosing different kinds of diseases, the optimum hyperparameter tuning of the DL models poses a major challenge to achieving maximum classification performance. To resolve these issues, this study develops a new Gene Expression Data Analysis using Artificial Intelligence for Prostate Cancer Diagnoses (GEDAAI-PCD) technique. The proposed GEDAAI-PCD technique examines the GED for the identification of PRC. To accomplish this, the GEDAAI-PCD technique initially normalizes the GED into a uniform format. In addition, the long short-term memory-deep belief network (LSTM-DBN) model was applied for PRC classification purposes. The wild horse optimization (EWHO) system was utilized as a hyperparameter tuning strategy to optimize the performance of the LSTM-DBN model. The experimental assessment of the GEDAAI-PCD system occurs on open open-accessed gene expression database. The experimental outcomes emphasized the supremacy of the GEDAAI-PCD method on PRC classification.


Medical Insurance Fraud Detection Based on Block Chain and Machine Learning Approach
Bijaya Kumar Sethi, Prakash Kumar Sarangi, and Adepu Sai Aashrith
IEEE
With the significant rise in medical costs, the Health Insurance Department's duty of controlling medical expenses has become increasingly vital. Traditional medical insurance settlements are paid per-service, which results in a lot of unnecessary costs. Now a day, the single-disease payment mechanism has been frequently employed to address this issue. However, there is a possibility of fraud with single-disease payments. In this work, the authors have presented a methodology for detecting the health insurance fraud entrenched block chain and Machine learning techniques like Support Vector Machine (SVM) and logistic Regression, that can automatically recognize apprehensive medical records to assure sustainable execution of single-disease payment and reduce medical insurance worker's workload. The authors have also proposed a medical record storage and management procedure based on consortium block chain to assure data security, immutability, traceability, and audit ability. The suggested system may effectively identify fraud and considerably increase the efficiency of medical insurance evaluations, as demonstrated by experiments on two real datasets from two 3A hospitals.


Medical Insurance Fraud Detection Based on Block Chain and Deep Learning Approach
Bijaya Kumar Sethi, Debabrata Singh, and Prakash Kumar Sarangi
IEEE
To control the medical expenses people are decided to do some insurance plans and the Health Insurance Department's duty of controlling medical expenses has become increasingly vital. Traditional medical insurance settlements are paid per-service, which results in a lot of unnecessary costs. Now a day, the single-disease payment mechanism has been frequently employed to address this issue. However, there is a possibility of fraud with single-disease payments. In this work, we have presented a methodology for detecting the health insurance fraud entrenched block chain and deep learning techniques, that can automatically recognize apprehensive medical records to assure sustainable execution of single-disease payment and reduce medical insurance worker's workload. We also proposed a medical record storage and management procedure based on consortium block chain to assure data security, immutability, traceability, and audit ability. The suggested system may effectively identify fraud and considerably increase the efficiency of medical insurance evaluations, as demonstrated by experiments on two real datasets from two hospitals.

RECENT SCHOLAR PUBLICATIONS

Prostate Cancer Prediction Using Convolutional Neural Networks
BK Sethi, D Singh, SK Rout
2024 IEEE North Karnataka Subsection Flagship International Conference 2024



Diabetes Prediction using Hybrid Machine Learning Techniques
BD Naik, A Spoorthy, S Baherji, SK Rout, BK Sethi
2024 OPJU International Technology Conference (OTCON) on Smart Computing for 2024



Breast Cancer Detection Using Convolutional Neural Network
SS Saniganti, SR Gaddam, SR Eppa, SK Rout, BK Sethi, B Kethireddy
2024 2nd International Conference on Disruptive Technologies (ICDT), 1409-1414 2024



A Precision Tessellated Fundus Detection: Leveraging Color and Texture Features with SVM Classification
K Anvesh, BM Reshmi, SK Rout, BK Sethi, SS Panigrahi
International Conference on Machine Learning, IoT and Big Data, 77-89 2024



Automated Interview Evaluation System Using RoBERTa Technology
GS Harsh, YSS Vivek, SK Rout, SR Reddy, BK Sethi
2024 1st International Conference on Cognitive, Green and Ubiquitous 2024



Mitigating Threats in PHY-Layer Authentication: A Proactive Defense Against Membership Inference Attacks in Wireless Signal Classifiers
D Madhuri, V Nikitha Reddy, MK Reddy, V Murthy, SK Rout, BK Sethi
International Conference on Broadband Communications, Networks and Systems 2024



Long Short-Term Memory-Deep Belief Network-Based Gene Expression Data Analysis for Prostate Cancer Detection and Classification
BK Sethi, D Singh, SK Rout, SK Panda
IEEE Access 12, 1508-1524 2023



Medical insurance fraud detection based on block chain and machine learning approach
BK Sethi, PK Sarangi, AS Aashrith
2022 Fourth International Conference on Emerging Research in Electronics 2022



Medical insurance fraud detection based on block chain and deep learning approach
BK Sethi, D Singh, PK Sarangi
2022 International Conference on Disruptive Technologies for Multi 2022



Detection: Leveraging Color and Texture
BK Sethi, SS Panigrahi
Intelligent Systems: Proceedings of 4th International Conference on Machine

MOST CITED SCHOLAR PUBLICATIONS

Long Short-Term Memory-Deep Belief Network-Based Gene Expression Data Analysis for Prostate Cancer Detection and Classification
BK Sethi, D Singh, SK Rout, SK Panda
IEEE Access 12, 1508-1524 2023
Citations: 22



Medical insurance fraud detection based on block chain and machine learning approach
BK Sethi, PK Sarangi, AS Aashrith
2022 Fourth International Conference on Emerging Research in Electronics 2022
Citations: 5



Medical insurance fraud detection based on block chain and deep learning approach
BK Sethi, D Singh, PK Sarangi
2022 International Conference on Disruptive Technologies for Multi 2022
Citations: 5



Breast Cancer Detection Using Convolutional Neural Network
SS Saniganti, SR Gaddam, SR Eppa, SK Rout, BK Sethi, B Kethireddy
2024 2nd International Conference on Disruptive Technologies (ICDT), 1409-1414 2024
Citations: 3