Intelligent Health Risk Prediction Framework Using a Hybrid Machine Learning Approach and LLM-Powered Feedback

The need for proactive and intelligent healthcare solutions has increased due to the ageing population, chronic disease incidence, and a requirement for real-time patient monitoring, particularly in rural areas. This paper presents an Intelligent Healthcare Monitoring System incorporating fall detection, natural language understanding, prediction of health risk through hybrid machine learning, real-time data gathering, and tracking of emergency events through GPS. An integrated model based on RF and DT is proposed with high precision for predicting and interpretation. RF improves the prediction accuracy, while DT makes decisions transparently. A weighted decision-making strategy is proposed that can be used for handling real-world noise in healthcare data. The model is built around 13 clinically significant features, stored using MQTT and NODE-RED. Preprocessing of the input data with noise removal, elimination of missing values, and data augmentation improves the data quality and generalization of the model. The proposed model is 97.93% accurate in its classification accuracy for the three health risk categories (low, medium, and high) at an 80:20 training-test split. The proposed system uses a chatbot based on Streamlit with a Large Language Model (LLM) to provide conversational, human-readable explanations for health predictions and provides personalized health advice. In the event of a fall or emergency, the system automatically communicates the patient’s GPS location to the caregivers or healthcare professionals so that they can respond quickly. It will help elderly care, personal healthcare, and rural healthcare by increasing access to medical care, patient safety, and emergency care using smart technologies.