Predicting Psychological Distress from Ecological Factors: A Machine Learning Approach

Over 300 million people worldwide were suffering from depression in 2017. Australia alone invests more than $9.1 billion each year on mental health related services. Traditional intervention methods require patients to first present with symptoms before diagnosis, leading to a reactive approach. A more proactive approach to this problem is highly desirable, and despite ongoing work using approaches such as machine learning, further work is required. This paper aims to provide a foundation by building a machine learning model across multiple techniques to predict psychological distress from ecological factors alone. Eight different classification techniques were implemented on a sample dataset, with the best results achieved through Logistic Regression, providing an accuracy of 0.811. The preliminary results suggest that, with future improvements to implementation and analysis, an accurate and reliable model is possible. This study, with the proposed base model, can potentially lead to the development of a proactive solution to the global mental health crisis.