An IoT-Based Framework for Personalized Health Assessment and Recommendations Using Machine Learning

To promote a healthy lifestyle, it is essential for individuals to maintain a well-balanced diet and engage in customized workouts tailored to their specific body conditions and health concerns. In this study, we present a framework that assesses an individual's existing health conditions, enabling people to evaluate their well-being conveniently without the need for a doctor's consultation. The framework includes a kit that measures various health indicators, such as body temperature, pulse rate, blood oxygen level, and body mass index (BMI), requiring minimal effort from nurses. To analyze the health parameters, we collected data from a diverse group of individuals aged 17-24, including both men and women. The dataset consists of pulse rate (BPM), blood oxygen level (SpO2), BMI, and body temperature, obtained through an integrated Internet of Things (IoT) unit. Prior to analysis, the data was augmented and balanced using machine learning algorithms. Our framework employs a two-stage classifier system to recommend a balanced diet and exercise based on the analyzed data. In this work, machine learning models are utilized to analyze specifically designed datasets for adult healthcare frameworks. Various techniques, including Random Forest, CatBoost classifier, Logistic Regression, and MLP classifier, are employed for this analysis. The algorithm demonstrates its highest accuracy when the training and testing datasets are divided in a 70:30 ratio, resulting in an average accuracy rate of approximately 99% for the mentioned algorithms. Through experimental analysis, we discovered that the CatBoost algorithm outperforms other approaches in terms of achieving maximum prediction accuracy. Additionally, we have developed an interactive web platform that facilitates easy interaction with the implemented framework, enhancing the user experience and accessibility.