AI-based Depression Detection using Profile Information

Depression is a severe mental health problem for people around the world, regardless of age, gender, or race. It is a cause of psychological disability, and these disorders can have an impact on a person’s interpersonal connections, such as work environment and family life, as well as their overall health and routines, such as irregular eating and sleeping patterns. However, unfortunately, the majority of cases of depression go undiagnosed and, therefore, untreated. Depression, when not detected at an earlier stage, can become a severe illness and may lead to suicide at later stages. Consequently, it becomes crucial to identify and prevent depression at an earlier stage. The data for this study are collected through a survey from undergraduates in consultation with psychiatrists and professors.Further, Natural Language Processing(NLP) techniques and Machine learning methodologies were used to train the data and evaluate the efficiency of the proposed model. This study looked at various feature selection (FS) techniques such as filter method Maximum Relevance and Minimum Redundancy-mRMR, wrapper method Recursive Feature Elimination-RFE, Boruta, Embedded method: Least Absolute Shrinkage and Selection Operator-LASSO were used to extract the most significant features from profile information of the user responsible for forming a depression. The Adaboost model produced an accuracy of 94% when considering all elements in the dataset. The different feature selection techniques, when applied, we found Adaboost with mRMR FS using Optuna Hypertuning produced an accuracy of 96%.