Ensemble-Driven Machine Learning Regression Models for Climate-Sensitive Crop Yield Prediction: A Comparative Performance Analysis Siva Subramanian R, Elumalai M, Saratha B, Ramesh K, Sudha K, Gnana Jeslin J Ssrg International Journal of Electronics and Communication Engineering, 2026 Precise forecasting of crop yields is the key to food security, resource management, and sustainable food farming. This paper will examine how different Machine Learning (ML) models can be used to predict crop yield in relation to climatic and other environmental conditions, like rainfall, temperature, and the use of pesticides. Multiple performance metrics, such as R², Mean Squared Error (MSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) were used to train and evaluate seven ML models which were Linear Regression (LR), Decision Tree (DT), K-Nearest Neighbors (KNN), Gradient Boosting (GB), XGBoost, Random Forest (RF), and Bagging. The experimental findings showed that the ensemble-based models were very effective compared to the traditional regression and distance-based algorithms. The Bagging recorded the best prediction accuracy in terms of R² score, closely followed by the RF. The two models were effective in capturing nonlinear relationships and high generalization in varied climatic and crop conditions. On the other hand, the simplicity of models like LR and KNN demonstrated low predictive abilities. The results highlight the scalability and the strength of the Ensemble Learning(EL) techniques in crop yield forecasting. The paper concludes with a set of recommendations on how to incorporate Explainable AI, real-time data that uses IoT, and region-specific hybrid deep learning systems to improve the interpretability, adjustment, and accuracy of agricultural forecasting systems in the future.
AN APPROACH TOWARDS DIABETIC RETINOPATHY DETECTION AND ANALYSIS THROUGH COGNITIVE COMPUTING B. Saratha, M S Radhika, Dr. V. Shenbaga Priya Archives for Technical Sciences, 2025 Diabetes is a common chronic condition that significantly impacts patients' daily lives. Although it cannot be cured, if left unmanaged, diabetes can progressively damage vital organs. Without early and appropriate care, it may lead to multiple adverse effects. To ensure proper care, diabetic individuals typically require regular visits to healthcare professionals. This study proposes a predictive method that empowers diabetic individuals to monitor and manage their blood sugar levels without frequent doctor visits. The central objective of the proposed approach is to reduce the dependence on physician consultations and diagnostic center appointments. To analyze diabetic retinopathy datasets, the proposed system employs Deep Predictive Neural Networks (DPNNs). Retinal lesions are identified using the Region Convergence Algorithm (RCA), and features are extracted using the Strong Intensity Extractor (SIE), which captures significant pixel-level information. Cognitive Computing (CC), integrated with DPNN, is applied to optimize classification accuracy. The model's performance is evaluated using metrics such as Accuracy, Precision, Recall, and the Confusion Matrix. Numerous experimental inputs are provided to the system based on the developed model to verify and predict potential abnormalities.
Future Horizons: Key Trends Shaping the Evolution of Machine Learning D. Prabhu, Tejasri M V S L, Durga Devi G, B. Saratha, Anish T P, Siva Subramanian R Proceedings of 8th International Conference on Trends in Electronics and Informatics Icoei 2025, 2025 Artificial intelligence - Machine learning (ML) is an emergent area of AI that aims to enhance the capability of a system to make predictions or decisions based on the data fed to it without the system being specifically programmed for such functions. Its importance stems from its capacity to analyze and abstract information, predict, and even make decisions in various contexts, transforming sectors such as healthcare, finance, and technology. With the increasingly large volume of data being created, the role played by ML in processing this data to discover meaningful patterns becomes more prominent. ML algorithms can be broadly categorized into three main types: supervised learning, unsupervised learning, and reinforcement learning. Linear regression and support vector machines, for instance, work from labeled training sets in order to make predictions and classifications. Such algorithms are used in many applications, like face recognition, filtering spam emails, and even diagnosing diseases. On the other hand, unsupervised learning operators, including K-means clustering and principal component analysis, work with data sets lacking labels to discover latent structures and patterns common in market segmentation and anomaly detection. Reinforcement learning can be viewed as enabling an agent to take a set of actions in an environment to maximize the total reward, as in Q-learning. This approach is especially useful in applications such as robotics, games, and the use of autonomous systems. In summary, machine learning algorithms are a key differentiator in improving productivity and fostering innovation, thereby facilitating advancements across various industries while creating future possibilities in artificial intelligence.
A Data-Driven Approach to Radio Frequency Signal Level Forecasting Using Machine Learning Algorithms M. Giridhar, V. Helen Deva Priya, S. Siamala Devi, B. Saratha, R. Jaidharni, S. Jeya Lakshmi Proceedings of the 9th International Conference on Inventive Systems and Control Icisc 2025, 2025 Advances in wireless technology allow autonomous wireless network deployments. Radiofrequency (RF). To integrate into networks, transmitters and receivers need to be aware of their environment and modify their broadcasting and receiving capacities. Because it can learn, evaluate, and forecast RF signals and environmental factors, machine learning is widely used. This dissertation tackles some of the challenges with RF learning approaches. Jamming and spoofing may render most machine learning algorithms useless when attackers are present. Adversarial learning is used to detect illegal RF spectrum use to allow learning in such circumstances. First, the researcher illustrates <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R F}$</tex> machine learning. Using separate cellular models, they build and deploy three recurrent neural networks for RF transmitter fingerprinting. Then safeguard dynamic spectrum access network broadcasts, which may be vulnerable to PUE assaults. A generative adversarial network (GAN) based solution to primary user emulation (PUE) attacks is proposed. Finally, recurrent neural network models predict principal users' DSA network activities so secondary users may exploit the shared spectrum opportunistically. Researchers use the specified learning models on testbeds utilizing Universal Software Radio Peripherals (USRPs) and Software Defined Radios (SDRs). Substantial improvements in the accuracy of RF transmitter characterization demonstrate the practical deployment capabilities of our models.
Green Data Centers: Advancing Sustainability in the Digital Era J. Elavarasi, G. Amudha, S.N. Ananthi, T. Thilagam, B. Saratha, Siva Subramanian R Proceedings of 5th International Conference on Trends in Material Science and Inventive Materials Ictmim 2025, 2025 The advancement of digital technologies, and the increased use of data-driven applications, have driven data center demand around the globe. Although crucial in contemporary architecture, conventional data centers are major energy hogs and detrimental to many environmental concerns including high carbon footprint and resource depletion. Green data centres are the new concept to counter the negative impacts of conventional data centres and these are more energy efficient, integrated with renewable power sources and use unique cooling technologies. These are: server virtualization, liquid and immersion cooling, waste heat to power recovery, and the use of clean energy such as solar and wind energy. AI driven power management, edge computing, and photonic interconnects are some of the new trends that continue to revolutionize the efficiency and sustainability of these facilities. However, challenges like high initial investment, policy instabilities, and technological limitations to renewable energy systems integration have not been addressed. Solving these problems should involve policy makers, managers and scholars in the field of supply chain management. Green data centres are the future in which, through the use of innovative technologies such as quantum computing, AI optimisation as well as modularity in the design of data centres, an increase in masses’ digital needs will be well served while at the same time being environmentally sensitive. With the help of such innovations, green data centers can set an example for the development of the new green economy.
RECENT SCHOLAR PUBLICATIONS
Deep Learning Applications in Predictive Analytics for Business Management JS Alikhan, B Sriman, R Jeyarani, E Venitha, B Saratha Mathematical Methods in Artificial Intelligence: Intelligent Systems, 303 , 2026 2026
Ensemble-Driven Machine Learning Regression Models for Climate-Sensitive Crop Yield Prediction: A Comparative Performance Analysis GJJ Siva Subramanian R, Elumalai M, B. Saratha, Ramesh K, Sudha Kothandapani International Journal of Electronics and Communication Engineering 13 (1) , 2026 2026
Enhancing Bird Flu Outbreak Predictions Using Data Mining Techniques K Deeba, K Jamberi, B Saratha International Conference on Advancements in Smart Computing and Information … , 2025 2025
A Data-Driven Approach to Radio Frequency Signal Level Forecasting Using Machine Learning Algorithms M Giridhar, VHD Priya, SS Devi, B Saratha, R Jaidharni, SJ Lakshmi 2025 9th International Conference on Inventive Systems and Control (ICISC … , 2025 2025
AN APPROACH TOWARDS DIABETIC RETINOPATHY DETECTION AND ANALYSIS THROUGH COGNITIVE COMPUTING. B Saratha, MS Radhika, VS Priya Archives for Technical Sciences/Arhiv za Tehnicke Nauke , 2025 2025 Citations: 2
Future Horizons: Key Trends Shaping the Evolution of Machine Learning D Prabhu, M Tejasri, B Saratha, TP Anish 2025 8th International Conference on Trends in Electronics and Informatics … , 2025 2025 Citations: 7
Green data centers: advancing sustainability in the digital era J Elavarasi, G Amudha, SN Ananthi, T Thilagam, B Saratha 2025 5th International Conference on Trends in Material Science and … , 2025 2025 Citations: 4
Enhancing Breast Cancer Detection with Machine Learning: A Predictive Modeling Approach NR J. Sherine Glory, M. Bhavani, B. Saratha, A. Akila, Dr. B. Prathusha ... South Eastern European Journal of Public Health , 2025 2025
AI-powered adaptive learning systems: Revolutionizing classroom education D Manoj, A Dutt, B Saratha Advancing Knowledge from Multidisciplinary Perspective Engineering … , 2024 2024 Citations: 3
MOST CITED SCHOLAR PUBLICATIONS
Future Horizons: Key Trends Shaping the Evolution of Machine Learning D Prabhu, M Tejasri, B Saratha, TP Anish 2025 8th International Conference on Trends in Electronics and Informatics … , 2025 2025 Citations: 7
Green data centers: advancing sustainability in the digital era J Elavarasi, G Amudha, SN Ananthi, T Thilagam, B Saratha 2025 5th International Conference on Trends in Material Science and … , 2025 2025 Citations: 4
AI-powered adaptive learning systems: Revolutionizing classroom education D Manoj, A Dutt, B Saratha Advancing Knowledge from Multidisciplinary Perspective Engineering … , 2024 2024 Citations: 3
AN APPROACH TOWARDS DIABETIC RETINOPATHY DETECTION AND ANALYSIS THROUGH COGNITIVE COMPUTING. B Saratha, MS Radhika, VS Priya Archives for Technical Sciences/Arhiv za Tehnicke Nauke , 2025 2025 Citations: 2
Deep Learning Applications in Predictive Analytics for Business Management JS Alikhan, B Sriman, R Jeyarani, E Venitha, B Saratha Mathematical Methods in Artificial Intelligence: Intelligent Systems, 303 , 2026 2026
Ensemble-Driven Machine Learning Regression Models for Climate-Sensitive Crop Yield Prediction: A Comparative Performance Analysis GJJ Siva Subramanian R, Elumalai M, B. Saratha, Ramesh K, Sudha Kothandapani International Journal of Electronics and Communication Engineering 13 (1) , 2026 2026
Enhancing Bird Flu Outbreak Predictions Using Data Mining Techniques K Deeba, K Jamberi, B Saratha International Conference on Advancements in Smart Computing and Information … , 2025 2025
A Data-Driven Approach to Radio Frequency Signal Level Forecasting Using Machine Learning Algorithms M Giridhar, VHD Priya, SS Devi, B Saratha, R Jaidharni, SJ Lakshmi 2025 9th International Conference on Inventive Systems and Control (ICISC … , 2025 2025
Enhancing Breast Cancer Detection with Machine Learning: A Predictive Modeling Approach NR J. Sherine Glory, M. Bhavani, B. Saratha, A. Akila, Dr. B. Prathusha ... South Eastern European Journal of Public Health , 2025 2025
