Prediction of Mental Health Problems Based on Logistic Regression Model

Abstract

This study aims to construct a logistic regression model using the "Mental Health Dataset" on Kaggle to predict mental health status and evaluate its accuracy based on the confusion matrix. The study selected 2,000 random samples and 16 variables, and after preprocessing the data, the model was built. Descriptive analysis found that "mental health history (binary)" was correlated to varying degrees with multiple variables, and ultimately 9 related factors were included in the model. The logistic regression results showed that social weaknesses, work interest, and mood swings were significantly related to the mental health status of college students. However, the model's predictive ability was limited, with an accuracy rate of 0.602, a Kappa value of -0.017, a sensitivity of only 0.053, and a specificity of 0.933. The study suggests that subsequent improvements in variable selection and model construction methods can enhance the predictive accuracy, providing a more reliable basis for early intervention and prevention of mental health problems.