Intrusion Detection in Wireless Sensor Networks Using Multitask Residual Shrinkage CNN Optimized by Fennec Fox Algorithm D. Satheesh Kumar, V. Niranjani, K. Pushpalatha, S. Gokila International Journal of Communication Systems, 2026 Wireless sensor networks (WSNs) are generally used in environmental monitoring, data transfer, and object detection but are susceptible to intrusion attacks based on their integration with the Internet of things (IoT). Most conventional intrusion detection systems experience high false alarm rates and excessive computational overhead. For better performance in overcoming these limitations, a new intrusion detection framework called Multitask Multiattention Residual Shrinkage Convolutional Neural Network with optimization through the Fennec Fox Optimization Algorithm (MMRSCNN‐FFOA‐ID‐WSN) is proposed. The framework combines preprocessing using an Ultrawideband Nanoplasmonic Bandpass Filter (UWNPBF) to eliminate redundant and biased entries, followed by feature selection through the Binary Waterwheel Plant Optimization Algorithm (BWWPOA). The MMRSCNN combines multitask learning with channel‐wise attention and residual shrinkage operations to detect attack types from input data. To further improve detection accuracy, the weight parameters are optimized using the FFOA, chosen for its faster convergence in high‐dimensional search spaces compared to other algorithms. Experiments were performed on the WSN‐DS dataset having normal traffic and various attack types such as flooding, black hole, gray hole, and time division multiple access attacks (TDMA). The proposed method achieved an accuracy of 99.46%, specificity of 98.83%, and a false alarm rate of 1.12%, which correspond to relative improvements of up to 22.37%, 25.32%, and 32.40%, respectively, over the baseline model. These results show that MMRSCNN‐FFOA‐ID‐WSN is an efficient and energy‐saving solution for WSNs intrusion detection.
Adaptive Machine Learning Framework for Comprehensive Disease Risk Forecasting From Electronic Health Record Premkumar Duraisamy, Abijith P Suthi, Ben Richards R, Niranjani V, J. Prasad, Karthik S Proceedings of the 4th International Conference on Intelligent Data Communication Technologies and Internet of Things Idciot 2026, 2026 This work investigates the application of machine learning algorithms for predicting the risk of multiple chronic diseases using Electronic Health Records (EHRs). Structured clinical datasets incorporating demographics, laboratory results, medical history, and lifestyle factors were utilized to evaluate six algorithms: Logistic Regression, Support Vector Classifier (SVC), K-Nearest Neighbors (KNN), Decision Tree, Random Forest, and Naïve Bayes. The models were assessed based on their accuracy, precision, and robustness in identifying disease risk patterns. Among them, the Random Forest algorithm achieved the highest accuracy of 94.4%, effectively capturing complex non-linear relationships across diverse patient profiles. Logistic Regression and SVC produced interpretable and stable outcomes, while KNN and Decision Tree demonstrated adaptability to heterogeneous data distributions. Despite its simplicity, Naïve Bayes achieved competitive performance in handling high-dimensional categorical variables. These findings highlight the potential of machine learning in enabling data-driven early disease risk prediction and advancing personalized healthcare. Future directions include integrating heterogeneous EHR data sources, leveraging deep learning techniques, and developing clinically interpretable decision support systems to enhance preventive care and improve patient outcomes.
Improved deep learning model for accurate energy demand prediction and conservation in electric vehicles integrated with cognitive radio networks V. Niranjani, Anandakumar Haldorai Scientific Reports, 2025 In the smart transport system, the immense growth of electric vehicles (EVs) and their charging demand is on the rise. However, the prediction of this demand has become a major issue. An increasing electrical vehicle number will result in a decrease the greenhouse gas releases. In the EV, the battery's capacity is limited and mileage anxiety is tedious. For the energy conservation of electric vehicles, many studies have been applied based on this concept. The problems addressed in existing research work are high in energy conservation. To overcome this issue, this paper proposed a model of Empirical Mode Decomposition with CNN and optimized with Seagull Optimization Algorithm (EMD-CNN-SOA). This proposed work provides an accurate prediction of demand for energy conservation and it reduces the burden on electric grids while minimizing the cost of charging. Cognitive radio (CR) in the form of wireless communication will revolutionize transportation through intelligent-based smart technology and it will anticipate the user needs in the aspects of detection of available bandwidths and frequencies then seamlessly connect the infrastructure and consumer devices. It will improve the safety of mobility and adapt to the current environmental situation, informing the driver about traffic congestion which saves energy. Cognitive radio sensors in the transportation will alert and measure the on-site real time conditions. The accuracy rate for the energy conservation in electric vehicles of TWC, LSTM 66.13%, Deep CNN 78.91%, RNN 83.46%, and proposed work of EMD-CNN-SOA 88.23%. Similarly, for CRN the accuracy rate of LSTM is 69.16%, Deep CNN is 86.25%, RNN is 84.37%, and the proposed work of EMD-CNN-SOA is 92.59%.
AI Driven Fiscal Planning through Budget Forecasting and Allocation using Time Series and Gradient Boosting Models Niranjani V, Shibikasri G, Swetha N, Thiksha Rubini S, Vishva Bharathi G 4th International Conference on Applied Artificial Intelligence and Computing Icaaic 2025, 2025 Effective budgetary distribution is one of the biggest challenges facing the government agencies because the conventional ways are majorly based on the past experience and lack consideration of the emerging economic and social dynamics. The present paper is based on the suggestion to develop an AI-based model to enhance the budgetary allocation in the Indian governmental sector with the use of Union Budget data (2016–2024), as well as of such macroeconomic variables as GDP, inflation, unemployment, and literacy rates. This system integrates time-series forecasting (ARIMA and Prophet) and machine learning forecasting (XGBoost and Random Forest) to learn the needs of the departmental spending, managers over/under-spending trends, and forms of spending in the future. To improve the accuracy of predictions, feature engineering uses lag variables, macroeconomic variables and the effects of the election year. The scenario simulation enables the policymakers to adjust the economic assumptions and test other budget results. The total budget is then assigned by an optimization module with reference to priority weights, policy goals and financial constraints. The results of experiments reveal the enhancement of prediction and decrease in inefficiencies. The system with an interactive dashboard is a holistic solution that improves transparency, accountability and efficiency in the management of public finances.
Implementation and Evaluation of Machine Learning Approaches for Parkinson’s Disease Prediction 16th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2025, 2025
Multi-Version Analysis of YOLO: A Deep Dive into YOLOv5, YOLOv7, and YOLOv9 for Face & Person Detection Syed Mohd Fazalul Haque, V Niranjani, Gotte Ranjith Kumar, Deepti Singh, Riya Kapoor, Sneh Kapoor 2025 7th International Conference on Information Systems and Computer Networks Iscon 2025, 2025 This study provides a comparative performance analysis of YOLOv5, YOLOv7, and YOLOv9 for real-time face and person detection, using key metrics like detection accuracy (mAP0.5), inference speed (FPS), and computational efficiency (GFLOPs). These models cater to critical applications such as autonomous vehicles, surveillance, and edge computing. YOLOv5 excels in lightweight deployments, achieving 91.2% mAP at 72 FPS. YOLOv7 improves accuracy to 93.6% at 58 FPS through enhanced feature aggregation. YOLOv9, the most advanced, achieves 95.1% mAP with multi-scale feature fusion and dynamic convolution, albeit at 41 FPS. The study highlights performance trade-offs and offers guidance for choosing models based on deployment needs and precision-critical requirements.
Machine Learning-Based Smart Stress Detection Using Sensor Data P. Kaushik, Md Ankushavali, Deepti Agrawal, Niranjani V, Chetan Khemraj, Viraj Gadde Proceedings 2025 IEEE 1st International Conference on Smart Innovations in Systems Infrastructure Mechanical Power AI and Computing Technologies Sisimpact 2025, 2025 In recent years, stress management has become an important area of research across various fields, including healthcare, workplace productivity, and mental well-being. The ability to detect stress levels accurately from physiological data can provide critical insights into an individual's emotional state, enabling timely interventions. In this research, we investigate state-of-the- art machine learning methods to effectively identify stress from sensor data. The data comprises multiple variables, including heart rate variability (HRV), electrodermal activity (EDA), skin temperature, and respiratory rate, among others. We focus on Support Vector Machines (SVM), a powerful classifier that excels in distinguishing patterns in high-dimensional spaces. In this context, SVM is utilized to classify stress levels (low, moderate, and high) based on sensor-derived physiological signals. By leveraging kernel techniques, SVM effectively separates stress levels using both linear and non-linear decision boundaries. The model's ability to generalize across unseen data makes it a strong candidate for real-time stress detection systems. To evaluate the model's effectiveness, we use metrics such as accuracy, precision, recall, and F1-score. Our findings suggest that SVM, with its robust classification capabilities, shows considerable promise in identifying stress patterns from sensor data. This has significant implications for fields such as healthcare and wearable technology, where stress monitoring can be integrated into devices to improve mental health support. Furthermore, our research demonstrates how SVM can enhance stress detection and personalized interventions, advancing the use of machine learning in emotion-sensing applications.
Implementation of a Machine Learning Model for Predicting and Evaluating Mental Health Risks 16th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2025, 2025
Real-Time Energy Optimization in Smart Buildings Using Predictive Edge Intelligence Deepali V Patil, Shripad Joshi, Niranjani V, Nagendar Yamsani, Thella Preethi Priyanka, Vinod Kumar Shukla 2025 International Conference on Intelligent and Secure Engineering Solutions Cises 2025, 2025 The increasing adoption of smart energy systems has revealed the constraints of centralized cloud architectures, especially concerning real-time data manipulation, energy optimization, and system modularity. In the era of IoT-enabled devices, the necessity for decentralized, efficient solutions is paramount. This study presents an AI- enabled edge computing architecture to enhance smart energy systems through distributed processing, predictive modeling, and autonomous control. The research facilitates the implementation of machine learning models at the network edge, and successfully minimizes latency, increases energy efficiency, and improves scalability when compared to traditional cloud architectures. The comparative study produced improved performance across multiple metrics, including response time, power consumption, and hands-free processing. Furthermore, fundamental challenges have been investigated, such as limited edge device resources, device heterogeneity, and synchronization of data devices. A proof-of- concept implementation has successfully demonstrated the feasibility of the architecture, despite limitations in hardware and device integration. Based on this work, further research will include efforts to further the reliability and sustainability of smart energy systems through real-time control, coordinated mechanisms, and federated learning to enable robust smart energy ecosystems across multiple platforms.
Pioneering Skin Cancer Prediction with Deep Learning Techniques 15th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2024, 2024
Block Chaining in Finance and Accounting V. Niranjani, V.S. Akshaya, V Harish, S Abhishek 2021 7th International Conference on Advanced Computing and Communication Systems Icaccs 2021, 2021
