A Machine-Learning Approach for Predicting Depression Through Demographic and Socioeconomic Features

According to the World Health organization, over 300 million people worldwide are affected by major depressive disorder (MDD). Individuals battling this serious mental condition may exhibit symptoms including anxiety, fatigue, and self-harm, all of which severely affect well-being and quality of life. Current trends in social media and population behavior bring up an urgent need for health professionals to strengthen mental health resources, improve access, and accurately diagnose depression. To mitigate the disparate impact of depression on different social and racial groups, this study identifies factors that strongly correlate with the prevalence of depression in the U.S., which focuses on adults using survey data from the 2019 pre-pandemic National Health Institute Survey (NHIS). In this study we trained four machine learning models—including Catboost, KNN, XGBoost—and concluded that a random forest model performed the best classification task on survey data of American adults, capable of an accuracy of 98.7%. Our results conclude that age, education, income, and household demographics are the primary factors impacting mental health. Awareness of these mental health stressors may help medical professionals, institutions, and governments to identify and diagnose more effectively the individuals at risk for MDD using clinical data. By receiving adequate mental health services, Americans can improve their quality of life and form a more fulfilling society.