Ponni Valavan M’s scientific contributions

Health monitoring has become a major concern in current society especially due to the continuously rising incidences of chronic ailments. Some of the primary monitoring techniques face a severe disadvantage in that they are incapable of identifying even the subtlest of health changes in real-time. Other systems also have difficulties handling large scales, nonlinear, dynamic data, for instance, from wearables' sensors. To deal with such problems, this work developed a new approach based on Autoencoder for learning the features without requiring the label and GAN for the online detection of the anomalous data in real-time health information. It involves wearables with IoT for perpetually monitoring the patients and using sophisticated detection algorithms for feasible cloud analysis. The initial performance shows a better accuracy/precision, and recall compared to the conventional models and serves as an efficient means to provide timely basic health intelligence.

Human-in-the-Loop (HITL) Machine Learning with Ensemble Learning uses the best of both automated algorithms and human input in making predictions and decision making when it comes to ML. In HITL frameworks they include human inputs in model training and validation, allowing the resultant models to be built based on actual needs rather than white algorithms, which may be insensitive to some details otherwise. It is more relevant in such areas or applications that involve higher risk such as credit risk applications in today’s financial world where predictions form the basis for major loans’ approval. Although previous studies that employed fully automated ML models have some shortcomings that include being prone to giving biased predictions, being less interpretable and scalability issues and failure to handle edge cases hence reducing reliability and trustworthiness. To overcome these problems, this study introduces a new approach of combining the HITL with the XGBoost, which is one of the most potent ensemble learning algorithms to develop a balanced model of loan default prediction. Through the analysis of loan data from Lending Club, this method has attained an accuracy of 99.4% along with a high precision and a high recall, thus pointing to well rounded performance of the model. Expert knowledge in the form of the HITL loop improves the regularization and optimization of the XGBoost model leading to a better, efficient, reliable, and more interpretable model for credit scoring. It was in python the model was developed taking advantage of the powerful ml and data analysis libraries which make it easy to develop, scalable and reusable for real life organizations especially in the financial arena where precision and accuracy are paramount. This approach lays some basis for subsequent innovations in the ICT applications in HITL ML systems in different fields.

The increasing demand for strong writing skills in academic and professional environments underscores the necessity of developing advanced feedback systems capable of delivering personalized and context-aware guidance. Traditional feedback methods often fail to provide the specific, timely, and individualized insights needed to enhance writing proficiency effectively. To address these limitations, this research introduces a novel approach utilizing the T5 model, a state-of-the-art Transformer-based architecture renowned for its performance across various language tasks. The proposed system leverages T5's text-to-text framework to offer comprehensive feedback on grammar correction, style enhancement, and overall writing improvement. By transforming written text into a format that the model can process and analyze, the T5 model enables the generation of feedback that is not only accurate but also tailored to the specific needs of the learner. The system's ability to provide real-time, contextually relevant feedback marks a significant advancement over conventional methods often lacking the depth and precision required for meaningful writing improvement. Experimental evaluations confirm the effectiveness of this approach, with the T5 model achieving a BLEU Score of 4.12, a ROUGE Score of 0.62, and a METEOR Score of 0.31. These metrics highlight the model's superior ability to align with reference texts and capture nuanced feedback, ultimately contributing to the enhancement of learners' writing skills. This research demonstrates the potential of the T5 model as a powerful tool in educational settings, offering a robust solution for personalized learning and fostering the development of advanced writing capabilities in diverse learning contexts.