Academic Depression Detection Using Behavioral Aspects for Sri Lankan University Students

Abstract

Academic Depression is a widespread problem among undergraduate students in Sri Lanka. It is exhausting and has a detrimental impact on students' academic life. Therefore, the development of a technique to estimate the probability of depression among undergraduates is a blessed respite. Depression is mostly caused by a failure to check students' psychological well-being on a regular basis. Identifying depression at the college level, leading the students to get proper therapy treatments. If a counselor detects depression in a student early enough, he/she can successfully assist the student in overcoming depression. However, keeping track of the substantial changes that occur in students because of depression becomes challenging for the counselor with a considerable number of undergraduates. The advancement of image processing and machine learning fields has contributed to the creation of effective algorithms capable of identifying depression probability. Depression Possibility Detection Tool (DPDT) is considered an effective automated tool that brings the depression probability of a certain student. In DPDT, the result is generated by concerning four main strategies. They are facial expressions, eye movements, behavior changes (step count and phone usage), and physical conditions (heart rate and sleep rate). Convolutional Neural Network (CNN) with Visual Geometry Group 16 (VGG16) model, Residual Neural Network (ResNet-50), Random Forest (RF) classifier is the main models and techniques used in the system. More than 93% of accuracy was generated in every trained model. The paper concludes the system overview along with four strategies, literature review, methodologies, conclusion, and future works.