The Research on Depression Prediction Among Student Based on Logistic Regression

Depression is one of most common mental illness, which influences human emotions and behavior. High proportion of students now are suffering from depression and predicting depression among students are becoming more and more significant in order to intervention treatment early. The student depression dataset is used to make a predictive model. Firstly, variables are selected by random forests. Next, the top four importance of variables are chosen to fit the logistic regression model. The four variables are having suicidal thoughts, academic pressure, financial stress and age. All of four variables contribute significantly to the model, which p-value are all less than 0.001. After that, the model is used in training data to evaluate performance and is compared with model with all variables. In this study, the AUC of model is good, at 0.905. However, the accuracy, sensitivity and specificity are 0.835, 0.780, 0.874, respectively, which are not high. Therefore, this model is not satisfying to predict depression among students. Multiplicate reasons could lead to low quality of model, such as variables selection. Further research could use other method to do variables selection or use other machine learning models to predict student depression.