Depression prediction among university students using deep learning techniques

Abstract

Depression is a health disorder which involves the affected individual experiencinglowmood which is the lack of interest or pleasure in usually pleasurable activities or events. It can greatly influence an individual’s emotions and health and the way she or he is capable of performing normal chores, participating in work or school, and establishingrelations with other people. Modern university students face the issue of depressionaffecting their performance and health. Therefore, the objective of this study is todiagnose depression among university students through deep learning. Participants’ information was gathered using a detailed questionnaire containing 1,029 entries and14variables: age, gender, semester, and responses to the questions formulated fromwell- standardized depression self-report instruments. For the prediction, we adopted RandomForest Classifier, Decision Tree Classifier, Voting Classifier, Gradient BoostingClassifier, and Artificial Neural Networks (ANN). The data was preprocessed inthecorrect way which includes; dealing with missing value issues, Normalizing the features, Categorical feature encoding. Simulations with the models were performed based on fundamental performance indicators, including accuracy, precision, recall rate, F1-score, and ROC-AUC. Based on the analysis, ANN gave the highest accuracy with 98.06%Theother methods which also gave a good prediction were, when the accuracy was at 97%there was Random Forest and Gradient Boosting.