RNN-LSTM BASED REGULAR HEALTH FACTOR ANALYSIS IN MEDICAL ENVIRONMENT

In an era where fast-paced routines, high stress, and unhealthy habits have become the norm, modern society is facing a surge in health problems such as high blood pressure, diabetes, poor sleep, and chronic stress. These lifestyle-related conditions often go unnoticed until they become severe, making early detection and preventive care more critical than ever. This project addresses that urgent need by leveraging advanced AI-powered deep learning models, specifically RNN and LSTM, to analyze commonly available health metrics—such as stress levels, blood pressure, glucose, cholesterol, and sleep duration— and accurately predict whether an individual is at high or low health risk. It empowers users to monitor their wellbeing regularly without relying on constant medical supervision, while also supporting healthcare professionals in identifying high-risk individuals who require timely attention. By shifting the focus from reactive treatments to proactive health management, the system promotes healthier lifestyles. This study presents a deep learning-based health risk prediction model using Bidirectional Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) networks to analyze regular health factors. The model focuses on early identification of individuals at risk by processing commonly available health attributes such as stress levels, blood pressure, glucose levels, cholesterol levels, sleep duration, activity level, heart rate, and sex. A structured data pipeline was followed, beginning with an 8-feature dataset of 982 records. Standardization and reshaping techniques were applied to prepare the data for sequential deep learning models. Unlike traditional machine learning approaches, this model utilizes the temporal learning capabilities of RNN and LSTM architectures to capture intricate, non-linear relationships between health parameters. Bidirectional layers further enhance accuracy by analyzing patterns in both forward and backward directions. The model was trained with early stopping and learning rate scheduling to prevent overfitting and improve convergence. The Bidirectional LSTM model achieved superior performance with a test accuracy of over 90%, outperforming the Simple RNN variant. Designed for scalability and realtime integration, the model provides a lightweight and accurate solution for personalized, preventive healthcare. This approach demonstrates the effectiveness deep learning models like RNN, LSTM for heart disease prediction, providing a practi-cal solution for early diagnosis and informed decision-making in medical practice.