College Student Stress Level Prediction with Ordinal Random Forest

College students nowadays seem to experience a more stressful life than decades ago, which makes monitoring the mental health of students necessary. This article aims to establish a prediction model, in the hope of early detecting students in need and performing proactive intervention. The data used in the project originates from Kaggle, a website, in the form of a Likert scale. To train a classifier for student stress level as the predictor, data are divided into a training dataset and a testing dataset, and variables are downsized using Kendall's rank correlation and ordinal logistic regression. Following that, reduced variables are put into an ordinal random forest to train the classifier. In the end, the classifier is tested against the test dataset, providing evaluators such as accuracy, precision, F1, etc. The F1 values of three target categories are 0.471 (No stress), 0.958 (Eustress), and 0.600 (Distress), respectively. The model obtained an overall macro-average F1 of 0.676, which is considered acceptable within the field of social and psychological sciences. Meanwhile, this value also suggests the prediction of “Eustress”; this stress level is better than the others.