Machine Learning with Human Resources Data: Predicting Turnover among Community Mental Health Center Employees.

Area of Science:

• Workforce Management
• Data Science in Healthcare
• Organizational Psychology

Background:

• Human Resources (HR) departments possess extensive employee data valuable for predicting turnover.
• Complex data formats often hinder the utilization of HR data for addressing employee turnover.
• Community mental health centers face challenges in leveraging their own data for retention strategies.

Purpose of the Study:

• To predict employee turnover in community mental health centers using machine learning (ML) on HR data.
• To evaluate the feasibility and accuracy of ML approaches for turnover prediction.
• To identify key HR data predictors of employee turnover.

Main Methods:

• Acquired historical HR data from two community mental health centers.
• Applied ML models, including random forest and lasso regression, for training and prediction.
• Assessed the feasibility of HR data extraction and processing for ML applications.

Main Results:

• The random forest model demonstrated strong predictive accuracy for employee turnover (Area Under the Curve > 0.8).
• ML methods identified significant turnover predictors such as past work years, wage, work hours, age, job position, training hours, and marital status.
• The process of extracting HR data for ML applications was found to be feasible.

Conclusions:

• ML approaches are feasible for predicting individual employee turnover using routinely collected HR data.
• These ML tools can identify employees at high risk of turnover, enabling targeted interventions.
• Organization-specific insights from ML can inform HR and leadership to address turnover and improve service quality.