Assessing Machine Learning Algorithms for Stress Prediction: A Healthcare Perspective

As stress related illnesses become more frequent, it is crucial that new techniques for reliable and quick forecasting are developed. This paper aims to analyse machine learning techniques with health data to measure stress level. The data collected from the Chengalpattu Medical Hospital has information of approximately 1000 entries with various parameters such as gender, seasonal cortisol level, sugar level, insulin level and tested HOMA-IR algorithm including Random forest(RF), Light gradient, Support Vector Classification (SVC), Logistic regression, K-nearest neighbors (KNN), Decision tree etc. Through the extensive comparison of these models, the LightGBM algorithm seems to have better accuracy. This study revealed that there is a possibility for high level machine learning to enhance the prediction of stress related-conditions thus enhancing the provision of health care and results will be advanced.