Predicting Depression Among University Students Using Machine Learning

Abstract

Depression is a major disorder and a growing problem that impacts a person’s way of living, disrupting natural functioning and impeding thought processes while they might remain oblivious to the fact that they are suffering from such a disease. Depression is especially prevalent in the younger population of underdeveloped and developing countries. Youth in countries such as Bangladesh face difficulties with studies, jobs, relationships, drugs, family problems which are all major or minor contributors in a pathway to depression. Furthermore, people in Bangladesh are not comfortable in speaking about this illness and often misinterpret this disorder as madness. This research, besides predicting depression in university undergraduates for the purpose of recommendation to a psychiatrist, focuses on gaining valuable insights as to why university students of Bangladesh, undergraduates in particular suffer from depression. The data for this research was collected by a survey designed after consultation with psychologists, counselors and professors. The survey was carried out through paper and Google survey form. The data was analyzed to find out relevant features related to depression using Random Forest Algorithm and then predict depression based on those features. A best method for predicting depression among Bangladesh undergraduates was found after using six algorithms to train and test the dataset. Deep Learning was found to be the best algorithm with the lowest number of false negatives, closely followed by Gradient Boost Algorithm both with an F-Measure of 63%. Generalized Linear Model, Random Forest, K-Nearest Neighbor and Support Vector Machine were the other four algorithms used for comparison. The objective of this research is to determine reasons for depression and to check whether depression can be successfully predicted with the help of related features. Depression is an illness that people in Bangladesh tend to ignore and hence it builds up and worsens with time. This research aims to identify depression in its early stages and ensure a fast recovery for victims so that heartbreaking incidents like suicide can be avoided.