Girija Attigeri
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Girija Attigeri has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Radhika M. Pai
Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, India
5 shared publications M. M. Manohara Pai
Department of Information and Communication Technology, Manipal Institute of Technology Manipal University, Manipal
2 shared publicationsManohara Pai M M
2 shared publications- A. Kotegar Karunakar
Department of Data Science and Computer Applications, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, India
1 shared publication A. V. Usharani
Department of Information and Communication Technology, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, Karnataka, India
1 shared publication- Abdul Tawab Akrami
Department of Information and Communication Technology, Manipal Academy of Higher Education, Manipal Institute of Technology, Manipal, India
1 shared publication
Latest Publications
- Open Access
- 20 October 2026
- Published by Springer Nature in Discover Computing
- Vol. 29 (1) , 777
Model-agnostics for real waste management using image classification
Automated waste segregation has emerged as a critical requirement for sustainable and safe urban waste management. While deep learning–based image classification models have demonstrated promising accuracy in waste identification, their black-box nature limits trust and reliability in real-world deployments. This paper presents a systematic evaluation study that combines the EfficientNet-B0 architecture with Local Interpretable Model-Agnostic Explanations (LIME) to investigate not only whether a waste classifier achieves high accuracy, but also whether its correct predictions are driven by semantically meaningful visual features rather than spurious background correlations. The proposed approach was initially benchmarked on the standard CIFAR-10 and CIFAR-100 datasets to establish baseline robustness, before being applied to the RealWaste dataset. The EfficientNet-based model achieves classification accuracies up to 93.08% on RealWaste, and across five random seeds, it achieves the highest mean test accuracy (91.30% $$\pm$$ 1.11%, vs. 90.04% $$\pm$$ 1.22% for ResNet50, 89.33% $$\pm$$ 1.13% for InceptionV3, and 86.02% $$\pm$$ 3.12% for MobileNetV2) while using fewer parameters than all models, except MobileNetV2. Beyond accuracy, LIME-based explanations analyze the validity of model decisions, revealing both meaningful feature utilization and instances of spurious visual correlations, supported by deletion/insertion AOPC scores and stability analysis. The study highlights the importance of combining predictive performance with explanation-driven evaluation and demonstrates the potential for explainable deep learning for future research on automated waste management. All results are based on controlled experimental conditions; scalability and real-world generalization require further validation.- 20 October 2026
- Published by Springer Nature in Discover Computing
- Vol. 29 (1) , 777
Open AccessModel-agnostics for real waste management using image classification
Automated waste segregation has emerged as a critical requirement for sustainable and safe urban waste management. While deep learning–based image classification models have demonstrated promising accuracy in waste identification, their black-box nature limits trust and reliability in real-world deployments. This paper presents a systematic evaluation study that combines the EfficientNet-B0 architecture with Local Interpretable Model-Agnostic Explanations (LIME) to investigate not only whether a waste classifier achieves high accuracy, but also whether its correct predictions are driven by semantically meaningful visual features rather than spurious background correlations. The proposed approach was initially benchmarked on the standard CIFAR-10 and CIFAR-100 datasets to establish baseline robustness, before being applied to the RealWaste dataset. The EfficientNet-based model achieves classification accuracies up to 93.08% on RealWaste, and across five random seeds, it achieves the highest mean test accuracy (91.30% $$\pm$$ 1.11%, vs. 90.04% $$\pm$$ 1.22% for ResNet50, 89.33% $$\pm$$ 1.13% for InceptionV3, and 86.02% $$\pm$$ 3.12% for MobileNetV2) while using fewer parameters than all models, except MobileNetV2. Beyond accuracy, LIME-based explanations analyze the validity of model decisions, revealing both meaningful feature utilization and instances of spurious visual correlations, supported by deletion/insertion AOPC scores and stability analysis. The study highlights the importance of combining predictive performance with explanation-driven evaluation and demonstrates the potential for explainable deep learning for future research on automated waste management. All results are based on controlled experimental conditions; scalability and real-world generalization require further validation. - Open Access
- 22 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 71408-71431
Comprehensive Review of Collaborative Data Caching in Edge Computing
Edge computing has gained significant attention due to the swift development of wireless communication technology. As more smart, data-intensive applications and Internet of Things (IoT) devices are used, the amount of data traffic grows at an exponential rate. This means that we need to manage data effectively so that services can be scalable, bandwidth-efficient, and latency-sensitive. Collaborative data caching has emerged as an essential technique in this context to meet the storage and retrieval needs of edge computing systems. We thoroughly examine the current collaborative caching techniques in edge computing, with a focus on recent developments and emerging approaches. We have categorized collaborative caching strategies into four main types: models based on stochastic, game theoretic, and mathematical methods to deal with uncertainty in network conditions and optimize resource allocation; machine learning-based models that employ artificial intelligence (AI) for content popularity prediction and cache optimization; heuristic models providing lightweight solutions for cache placement and replacement; and hybrid models combining multiple strategies. Each of these models is intended to maximize performance under various network and data conditions. We conclude by looking into some of the open problems and difficulties with edge caching to promote further research in this field.
- 22 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 71408-71431
Open AccessComprehensive Review of Collaborative Data Caching in Edge Computing
Edge computing has gained significant attention due to the swift development of wireless communication technology. As more smart, data-intensive applications and Internet of Things (IoT) devices are used, the amount of data traffic grows at an exponential rate. This means that we need to manage data effectively so that services can be scalable, bandwidth-efficient, and latency-sensitive. Collaborative data caching has emerged as an essential technique in this context to meet the storage and retrieval needs of edge computing systems. We thoroughly examine the current collaborative caching techniques in edge computing, with a focus on recent developments and emerging approaches. We have categorized collaborative caching strategies into four main types: models based on stochastic, game theoretic, and mathematical methods to deal with uncertainty in network conditions and optimize resource allocation; machine learning-based models that employ artificial intelligence (AI) for content popularity prediction and cache optimization; heuristic models providing lightweight solutions for cache placement and replacement; and hybrid models combining multiple strategies. Each of these models is intended to maximize performance under various network and data conditions. We conclude by looking into some of the open problems and difficulties with edge caching to promote further research in this field.
- Open Access
- 22 October 2024
- Published by Springer Nature in International Journal of Information Technology
- p. 1-16
Optimization model for mass casualty management system using qos-aware routing protocol and casualty triage prediction
Purpose: Unexpected events, whether man-made or natural, cause significant distress and disorder. The number and magnitude of disasters and catastrophes have been rising globally in recent decades, according to historical data. Continued real-time monitoring of mass casualty along with the arrangement of needed medical resources to handle the mass casualty is required to reduce the mortality and morbidity rates. Methods: Electronic tag-based casualty monitoring and machine learning-based Decision Support Systems are emerging as a more effective and proactive solution that provides continuous remote monitoring of patients. A novel framework based on Body-to-Body Network, Prediction model, and Genetic Algorithm-based medical resource optimization is proposed for the continuous monitoring of the mass casualty and medical resource allocation at the incident. The aim of this work is to give priority to the handling of critical casualties. Firstly, a Quality of Service and load-sensitive routing protocol for transmitting mass casualties' physiological parameters across a wireless network is proposed, with the critical casualty being emphasized. Secondly, the clinical seriousness degree of the mass casualty is predicted using Backpropagation Artificial Neural Network. Finally, an optimization model using a Genetic algorithm and queuing theory is proposed to find the required optimal number of medical resources to handle critical and non-critical casualties separately. Also, the proposed optimization model considers the predicted clinical level transition rates of the mass casualty. Results: The performance and accuracy of the proposed framework are evaluated using the MIMIC-II dataset. The outcome demonstrates that the framework emphasizes critical casualty management. Furthermore, the framework allocates an adequate number of servers by incorporating the proposed routing protocol in comparison to the AODV protocol. Conclusion: The inclusion of a prediction model in the framework aids in allocating an adequate number of servers by considering the predicted clinical deteriorating transition rates of casualties at a mass casualty incident. In terms of the estimated length of the casualty at the incident, the results suggest that incorporating a medical resource optimization model outperforms the non-optimal option.
- 22 October 2024
- Published by Springer Nature in International Journal of Information Technology
- p. 1-16
Open AccessOptimization model for mass casualty management system using qos-aware routing protocol and casualty triage prediction
Purpose: Unexpected events, whether man-made or natural, cause significant distress and disorder. The number and magnitude of disasters and catastrophes have been rising globally in recent decades, according to historical data. Continued real-time monitoring of mass casualty along with the arrangement of needed medical resources to handle the mass casualty is required to reduce the mortality and morbidity rates. Methods: Electronic tag-based casualty monitoring and machine learning-based Decision Support Systems are emerging as a more effective and proactive solution that provides continuous remote monitoring of patients. A novel framework based on Body-to-Body Network, Prediction model, and Genetic Algorithm-based medical resource optimization is proposed for the continuous monitoring of the mass casualty and medical resource allocation at the incident. The aim of this work is to give priority to the handling of critical casualties. Firstly, a Quality of Service and load-sensitive routing protocol for transmitting mass casualties' physiological parameters across a wireless network is proposed, with the critical casualty being emphasized. Secondly, the clinical seriousness degree of the mass casualty is predicted using Backpropagation Artificial Neural Network. Finally, an optimization model using a Genetic algorithm and queuing theory is proposed to find the required optimal number of medical resources to handle critical and non-critical casualties separately. Also, the proposed optimization model considers the predicted clinical level transition rates of the mass casualty. Results: The performance and accuracy of the proposed framework are evaluated using the MIMIC-II dataset. The outcome demonstrates that the framework emphasizes critical casualty management. Furthermore, the framework allocates an adequate number of servers by incorporating the proposed routing protocol in comparison to the AODV protocol. Conclusion: The inclusion of a prediction model in the framework aids in allocating an adequate number of servers by considering the predicted clinical deteriorating transition rates of casualties at a mass casualty incident. In terms of the estimated length of the casualty at the incident, the results suggest that incorporating a medical resource optimization model outperforms the non-optimal option.
- Open Access
- 18 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 156572-156587
Optimizing Question Answering Systems in Education: Addressing Domain-Specific Challenges
Question Answering (QA) systems are increasingly essential in educational institutions, enhancing both learning and administrative processes by providing quick and accurate answers to user queries. However, existing systems often struggle with accurately classifying and responding to diverse and context-dependent questions, especially when dealing with large knowledge graphs. Predicting the domain of a question can significantly narrow down the search space within a vast knowledge graph, improving the system’s efficiency and accuracy. This study addresses this gap by developing and evaluating domain prediction models. We compare the performance of various deep learning architectures, including Bi-GRU, Bi-LSTM, GRU, and LSTM. Our results demonstrate that the 1-layer Bi-GRU model outperforms the others, achieving the highest test accuracy of 82.13%. Additionally, by employing an ensemble technique that combines models with highest performance measures from each architecture, we further enhance overall performance, achieving an accuracy of 87.14%, which demonstrates improved predictive capability. This work is significant as it provides a robust solution for improving the accuracy and relevance of QA systems in educational settings, thereby enhancing user satisfaction and operational efficiency.
- 18 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 156572-156587
Open AccessOptimizing Question Answering Systems in Education: Addressing Domain-Specific Challenges
Question Answering (QA) systems are increasingly essential in educational institutions, enhancing both learning and administrative processes by providing quick and accurate answers to user queries. However, existing systems often struggle with accurately classifying and responding to diverse and context-dependent questions, especially when dealing with large knowledge graphs. Predicting the domain of a question can significantly narrow down the search space within a vast knowledge graph, improving the system’s efficiency and accuracy. This study addresses this gap by developing and evaluating domain prediction models. We compare the performance of various deep learning architectures, including Bi-GRU, Bi-LSTM, GRU, and LSTM. Our results demonstrate that the 1-layer Bi-GRU model outperforms the others, achieving the highest test accuracy of 82.13%. Additionally, by employing an ensemble technique that combines models with highest performance measures from each architecture, we further enhance overall performance, achieving an accuracy of 87.14%, which demonstrates improved predictive capability. This work is significant as it provides a robust solution for improving the accuracy and relevance of QA systems in educational settings, thereby enhancing user satisfaction and operational efficiency.
- Open Access
- 20 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 29633-29647
Advanced NLP Models for Technical University Information Chatbots: Development and Comparative Analysis
The process to obtain the information about university/institute is a critical stage in the academic journey of prospective students who are seeking information about the specific courses which makes that university/institute unique. This process begins with exploration to general information about universities through websites, rankings, and brochures from various sources. Most of the time, information available on different sources leads to discrepancies and influences student’s decisions. By addressing inquiries promptly and providing valuable information, universities can guide individuals in making informed choices about their academic future. To address this, the chatbot application is the most effective tool to be implemented and make it functional on university’s functional website. A chatbot is an artificially intelligent tool which can interact with humans and can mimic a conversation. This tool can be implemented using advanced Natural Language Processing (NLP) models to provide the pre-defined answers to the student’s queries. Chat-bot is very helpful for query resolution during the counseling process of the institute as it will provide official/uniform information and can be accessed 24x7. Therefore, the aim of this research work was to implement a chat-bot using various NLP models and compare them to identify best one. In this work, five chat-bot models were implemented using neural networks, TF-IDF vectorization, sequential modeling and pattern matching. From the results, it was observed that neural network-related models had better accuracy than TF-IDF and pattern matching model, and sequential modeling is the most accurate model because it prevents over-fitting. Furthermore, a chat-bot having any kind of optimizer can improve the result and it is most important that pattern matching, and semantic analysis should be the parts of a chat-bot for real time scenarios.
- 20 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 29633-29647
Open AccessAdvanced NLP Models for Technical University Information Chatbots: Development and Comparative Analysis
The process to obtain the information about university/institute is a critical stage in the academic journey of prospective students who are seeking information about the specific courses which makes that university/institute unique. This process begins with exploration to general information about universities through websites, rankings, and brochures from various sources. Most of the time, information available on different sources leads to discrepancies and influences student’s decisions. By addressing inquiries promptly and providing valuable information, universities can guide individuals in making informed choices about their academic future. To address this, the chatbot application is the most effective tool to be implemented and make it functional on university’s functional website. A chatbot is an artificially intelligent tool which can interact with humans and can mimic a conversation. This tool can be implemented using advanced Natural Language Processing (NLP) models to provide the pre-defined answers to the student’s queries. Chat-bot is very helpful for query resolution during the counseling process of the institute as it will provide official/uniform information and can be accessed 24x7. Therefore, the aim of this research work was to implement a chat-bot using various NLP models and compare them to identify best one. In this work, five chat-bot models were implemented using neural networks, TF-IDF vectorization, sequential modeling and pattern matching. From the results, it was observed that neural network-related models had better accuracy than TF-IDF and pattern matching model, and sequential modeling is the most accurate model because it prevents over-fitting. Furthermore, a chat-bot having any kind of optimizer can improve the result and it is most important that pattern matching, and semantic analysis should be the parts of a chat-bot for real time scenarios.
- 7 November 2022
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
- Vol. 36 (14)
Performance Analysis of Federated Learning Aggregation Algorithms for Secure and Efficient Data Handling
Traditional machine learning projects have revolved around training the model with the help of previously observed data to be able to predict output for future unknown data. In the current scenario, when the data generated are huge, centralized training of the model becomes inefficient. Hence, distributed approach with client server model has to be used for training the models. This introduces data handling and critical data privacy issues. This paper concentrates on Federated learning (FL) which builds a model for the server by aggregating the parameters obtained from the local models of the client devices. The research work focuses on design and evaluation of three new FL algorithms against the average of the performances of the local models. The evolved approach considering weights of the local models proportional to accuracy of the local model is found to be the most accurate and better than the centralized approach. The evaluation is done using three different algorithms belonging to regression and classification on multiple datasets. It is observed that there is only one round of communication between the clients and server required in the federated learning setup to achieve the benchmarked accuracy set by the centralized setup. This is a considerable development and state-of-the-art approach to reduce communication and computation costs.
- 7 November 2022
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
- Vol. 36 (14)
Performance Analysis of Federated Learning Aggregation Algorithms for Secure and Efficient Data Handling
Traditional machine learning projects have revolved around training the model with the help of previously observed data to be able to predict output for future unknown data. In the current scenario, when the data generated are huge, centralized training of the model becomes inefficient. Hence, distributed approach with client server model has to be used for training the models. This introduces data handling and critical data privacy issues. This paper concentrates on Federated learning (FL) which builds a model for the server by aggregating the parameters obtained from the local models of the client devices. The research work focuses on design and evaluation of three new FL algorithms against the average of the performances of the local models. The evolved approach considering weights of the local models proportional to accuracy of the local model is found to be the most accurate and better than the centralized approach. The evaluation is done using three different algorithms belonging to regression and classification on multiple datasets. It is observed that there is only one round of communication between the clients and server required in the federated learning setup to achieve the benchmarked accuracy set by the centralized setup. This is a considerable development and state-of-the-art approach to reduce communication and computation costs.
- Open Access
- 22 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 34404-34414
Secure EMR Classification and Deduplication Using MapReduce
Healthcare providers generate huge amount of data every day through registration, lab results, prescriptions, and others. This is stored in the form of Electronic Medical Records (EMR) in a central repository. A medical record data is very huge, difficult to read and understand. To give an insight to the professionals in analysing the different domains a patient belongs to, it is necessary to get pointers to a file before classifying it to a particular department for further analysis. This study provides a EMR processing system to automatically classify EMRs based on the important medical terms using TF-IDF and topic modelling. Automatic Classification of EMRs help the healthcare professionals in taking accurate decisions, providing efficient service, and improves the time taken for processing huge amount of data and in better organizing of patient. The data stored on the cloud may contain duplicate copies of EMR on several storage systems at file level thus increasing the network bandwidth, cost, and consuming storage space. Hence, a deduplication mechanism is required to avoid or reduce the data redundancy. Adapting cloud computing for healthcare systems necessitates sharing patient data with cloud service providers, which creates security concerns as the data may contain diagnosis, medication, laboratory results and medical claims. The main aim of this work is to classify the EMRs as per the specialization using KNN algorithm, optimize storage using deduplication and protect the data using DNA encryption algorithm before uploading to Hadoop. Data redundancy is taken care by implementing deduplication techniques using MD5 hashing.
- 22 March 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 34404-34414
Open AccessSecure EMR Classification and Deduplication Using MapReduce
Healthcare providers generate huge amount of data every day through registration, lab results, prescriptions, and others. This is stored in the form of Electronic Medical Records (EMR) in a central repository. A medical record data is very huge, difficult to read and understand. To give an insight to the professionals in analysing the different domains a patient belongs to, it is necessary to get pointers to a file before classifying it to a particular department for further analysis. This study provides a EMR processing system to automatically classify EMRs based on the important medical terms using TF-IDF and topic modelling. Automatic Classification of EMRs help the healthcare professionals in taking accurate decisions, providing efficient service, and improves the time taken for processing huge amount of data and in better organizing of patient. The data stored on the cloud may contain duplicate copies of EMR on several storage systems at file level thus increasing the network bandwidth, cost, and consuming storage space. Hence, a deduplication mechanism is required to avoid or reduce the data redundancy. Adapting cloud computing for healthcare systems necessitates sharing patient data with cloud service providers, which creates security concerns as the data may contain diagnosis, medication, laboratory results and medical claims. The main aim of this work is to classify the EMRs as per the specialization using KNN algorithm, optimize storage using deduplication and protect the data using DNA encryption algorithm before uploading to Hadoop. Data redundancy is taken care by implementing deduplication techniques using MD5 hashing.
- Open Access
- 1 October 2019
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 9 (5) , 3786-3797
Framework to predict NPA/Willful defaults in corporate loans: a big data approach
Growth and development of the economy is dependent on the banking system. Bad loans which are Non-Performing Assets (NPA) are the measure for assessing the financial health of the bank. It is very important to control NPA as it affects the profitability, and deteriorates the quality of assets of the bank. It is observed that there is a significant rise in the number of willful defaulters. Hence systematic identification, awareness and assessment of parameters is essential for early prediction of willful default behavior. The main objective of the paper is to identify exhaustive list of parameters essential for predicting whether the loan will become NPA and thereby willful default. This process includes understanding of existing system to check NPAs and identifying the critical parameters. Also propose a framework for NPA/Willful default identification. The framework classifies the data comprising of structured and unstructured parameters as NPA/Willful default or not. In order to select the best classification model in the framework an experimentation is conducted on loan dataset on big data platform. Since the loan data is structured, unstructured component is incorporated by generating synthetic data. The results indicate that neural network model gives best accuracy and hence considered in the framework.
- 1 October 2019
- Published by Institute of Advanced Engineering and Science in International Journal of Electrical and Computer Engineering (IJECE)
- Vol. 9 (5) , 3786-3797
Open AccessFramework to predict NPA/Willful defaults in corporate loans: a big data approach
Growth and development of the economy is dependent on the banking system. Bad loans which are Non-Performing Assets (NPA) are the measure for assessing the financial health of the bank. It is very important to control NPA as it affects the profitability, and deteriorates the quality of assets of the bank. It is observed that there is a significant rise in the number of willful defaulters. Hence systematic identification, awareness and assessment of parameters is essential for early prediction of willful default behavior. The main objective of the paper is to identify exhaustive list of parameters essential for predicting whether the loan will become NPA and thereby willful default. This process includes understanding of existing system to check NPAs and identifying the critical parameters. Also propose a framework for NPA/Willful default identification. The framework classifies the data comprising of structured and unstructured parameters as NPA/Willful default or not. In order to select the best classification model in the framework an experimentation is conducted on loan dataset on big data platform. Since the loan data is structured, unstructured component is incorporated by generating synthetic data. The results indicate that neural network model gives best accuracy and hence considered in the framework.
- 23 August 2018
- Published by Springer Nature
- p. 177-187
Comparative Analysis of Prediction Algorithms for Diabetes
- 23 August 2018
- Published by Springer Nature
- p. 177-187
Comparative Analysis of Prediction Algorithms for Diabetes
- Open Access
- 1 January 2018
- Published by Elsevier in Procedia Computer Science
- Vol. 135, 369-376
Knowledge Base Ontology Building For Fraud Detection Using Topic Modeling
- 1 January 2018
- Published by Elsevier in Procedia Computer Science
- Vol. 135, 369-376
Open AccessKnowledge Base Ontology Building For Fraud Detection Using Topic Modeling