Depression, anxiety, and stress among Indonesians during COVID-19 pandemic: Multiclass prediction machine learning approach

The COVID-19 pandemic in Indonesia has recorded an increase in stress levels of around 75% in 2021. This problem shows the importance of further testing to classify a person’s condition based on the level of depression, anxiety, and stress (DAS). A good classification requires the latest methods in its approach. Therefore, this study aims to compare suitable machine learning approaches to predict the level of DAS. This research presents nine machine-learning approaches to classify a person’s category based on the dimensions of DAS. Logistic regression, support vector classification (SVC), k-nearest neighbors, gaussian naïve Bayesian, random forest, stochastic gradient descent, linear SVC, gradient boosting, and decision tree were applied to primary data from a survey of 344 people who completed the depression, anxiety, and stress questionnaire. This study found that the two best approaches, based on the performance of the cross-validation mean scores metric, were logistic regression and support vector classification. Both methods provided an accuracy value of >85% in the classification of each dimension. In addition, other explanations were found from a health perspective. The findings in this study can be used as reference material for dealing with stress during a pandemic or post-pandemic to achieve normal conditions.