JivaCare: A Smart Health Care Application Integrating Home Remedies for Holistic Well-Being

JivaCare is a smart health care application that aims to diagnose common illnesses and recommend home remedies by leveraging advancements in Artificial Intelligence (AI) and Machine Learning (ML). Initially, a custom dataset of diseases and symptoms was used to develop a disease-diagnoses chatbot. However, challenges such as limited accuracy and rigid disease-symptom mapping were identified. To address these, GPT-3.5 API was integrated via RapidAPI, enhancing conversational quality without requiring extensive training while improving response quality. A MySQL database was implemented to store conversational history and session-based memory. Subsequently, the focus transitioned to a text classification approach using a 5,634-sample dataset from HuggingFace. This enabled flexible symptom-to-disease classification, overcoming the limitations of the initial dataset. Five machine learning models were evaluated, with Logistic Regression achieving the highest accuracie of 85 % after fine-tuning its hyperparameters. Additionally, neural network architectures such as GRU, RNN and CNN were also employed, achieving validation accuracies of 76%, 82% and 84%, respectively. The results demonstrate the effectiveness of integrating ML and deep learning techniques for accurate disease prediction and remedy recommendation. This work can establish the foundation for a scalable and user-friendly healthcare system, bridging the gap between AI and personalised natural health benefits.