Automated Machine Learning Algorithms for Predicting Anxiety and Depression in Bangladeshi University Students

Abstract

Stress is a mental health issue that results in a persistent sense of hopelessness and boredom. It affects one's thoughts, feelings, and behavior and sets off a host of psychological and medical problems. The global rate of increase in mental stress is unparalleled, particularly among students. This problem is caused by a number of factors, and the infection is causing an increase in linked illnesses. Not only does gloom increase the likelihood of health hazards, but it can also lead to dangerous social offenses such as self-harm and abuse within the family. We employed machine learning tools such as logistic regression (LR), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Adaboost, and Bagging to forecast psychological stress or anxiety and determined the correlation heat map for observing related features. We have attempted to look at eight machine learning algorithms concerning performance time and accuracy. In our proposed framework RF, DT, and Adaboost show 100% accuracy and precision, but in the perspective of performance time, SVM is the best because it takes only 0.007 seconds. The primary objective of this study is to forecast the anxiety and depression disorders of Bangladeshi university students using a machine learning algorithm. We assessed the performance of eight different machine learning algorithm compositions using a 10-fold verification technique. Random forest classifiers typically outperform other machine learning classifiers in terms of performance.