Machine learning models for predicting depression in Korean young employees.

Area of Science:

• Occupational Health
• Computational Psychiatry
• Data Science in Healthcare

Background:

• Rising incidence of depression among employees.
• Limitations of traditional statistical methods in predicting workplace depression.
• Need for advanced analytical approaches to identify depression risk factors.

Purpose of the Study:

• To apply machine learning algorithms for detecting depression risk in employees.
• To identify key factors associated with workplace depression.
• To compare the predictive performance of different machine learning models.

Main Methods:

• Utilized a dataset of 503 employees with 27 predictor variables.
• Employed sparse logistic regression, support vector machine, and random forest models.
• Evaluated models based on accuracy, precision, sensitivity, specificity, and AUC.

Main Results:

• Random forest achieved the highest accuracy (88.7%), while sparse logistic regression and support vector machine showed 86.8% accuracy.
• Identified significant factors: gender, physical health, job-related aspects, and psychosocial risk/protective factors.
• Machine learning models demonstrated comparable predictive performance.

Conclusions:

• Machine learning models show strong potential for predicting employee depression risk.
• Identified factors can inform the development of intelligent mental healthcare systems.
• Early detection of depressive symptoms in the workplace is achievable through these methods.