A study of Machine Learning Methods Based Affective Disorder Detection using Multi-Class Classification

Abstract

Psychological health issues like stress, anxiety and depression is becoming general due to various factors. Different severity levels of stress, anxiety and depression have varied impacts on people, which can lead to suicidal ideation and suicide attempts. Demographic data consists of gender, age group, marital status, education, type of employment, economic status, and living status, and 21 questions about stress, anxiety, and depression, including scores, were considered for the dataset. The proposed K-nearest neighbor model achieved maximum accuracy with 94.5% for stress and anxiety and 97.7% for depression, compared with Naive Bayes's 81.8%, 81.5%, and 83.5% as minimum accuracy for stress, anxiety, and depression, respectively, as compared with other models.