Predicting and Monitoring Anxiety and Depression: Advanced Machine Learning Techniques for Mental Health Analysis

Abstract

Anxiety and depression are considered among the most prevailing mental illnesses; they affect millions in the USA and worldwide. Besides being highly prevalent, these conditions have major implications for individuals and American society as a whole. The prime objective of this research project was to design and evaluate advanced machine learning methodologies for the monitoring and prediction of anxiety and depression. The rise in recent advances in Machine Learning and AI technologies has unleashed tremendous potential in the diagnosis and monitoring of mental health conditions such as anxiety and depression. Predictive models, powered by Machine Learning algorithms, process vast amounts of data and detect patterns that might have evaded human clinicians. This dataset for the current research project was retrieved from the website kaggle.com and shared publicly with anyone by the Harvard Data Verse repository. The dataset contained behavioral, psychophysiological, and demographic data that were collected from 593 participants aged 18-35 years for the prediction of anxiety and depression disorder risk. For this study, three machine learning algorithms were deployed: Logistic Regression, XG-Boost, and Random Forest. To assess and evaluate the performance of the algorithms, two key performance evaluation metrics were utilized MSE & R-squared. By reviewing the performance of the aforementioned three machine learning models, Linear Regression, Random Forest Regressor, and XG-Boost Regressor, using evaluation metrics MSE and R-squared are compared in a tabular form. Retrospectively, all three models performed remarkably well, with very low MSE values and R-squared values close to 1. Linear Regression marginally outperformed the others, but all models were successful in predicting the anxiety or depression indicator accurately. The proposed models are valid and reliable models for predicting mental health, therefore enabling the identification of at-risk individuals well in advance, allowing early intervention to prevent symptom onsets or advancements in their course and thus improve overall outcomes.