Identifying risk factors for depression and positive/negative mood changes in college students using machine learning

Area of Science:

• Psychology
• Computer Science
• Mental Health

Background:

• College students face significant mental health challenges, including depression.
• Understanding predictors of depression changes is crucial for timely intervention.

Purpose of the Study:

• To apply machine learning models to predict the magnitude of depression changes in college students.
• To identify key psychological variables influencing depression fluctuations.

Main Methods:

• Collected data from college students on depression, demographics, parenting styles, mental health, personality, coping, SCL-90, and social support.
• Utilized logistic regression, random forest, support vector machine (SVM), and k-nearest neighbor algorithms.
• Selected the best-performing model and analyzed feature importance.

Main Results:

• Support Vector Machines (SVM) demonstrated superior performance, achieving 89.4% accuracy for predicting negative depression changes and 91.9% for positive changes.
• Baseline depression levels, father's emotional expression, and mother's emotional expression were identified as significant predictors.
• These factors were important for predicting both increases and decreases in depression.

Conclusions:

• Machine learning models effectively predict the extent of depression changes in college students.
• Parental emotional expression and initial depression levels are critical factors in predicting depression trajectories.
• Findings offer novel approaches for psychological health research and clinical practice.