An Objective-Based Entropy Approach for Interpretable Decision Tree Models in Support of Human Resource Management:...

Area of Science:

• Human Resource Management
• Data Science
• Organizational Behavior

Background:

• Employee absenteeism negatively impacts organizational productivity and profitability.
• Understanding absenteeism causes and identifying at-risk employee subgroups is crucial for mitigation.
• Traditional statistical models often identify only simple correlations, limiting deeper insights.

Purpose of the Study:

• To apply interpretable classification algorithms for uncovering employee subgroups with shared characteristics and absenteeism levels.
• To assist human resource managers in understanding the root causes of absenteeism.
• To demonstrate the value of interpretable models in supporting human resource decision-making.

Main Methods:

• Utilized an objective-based information gain measure.
• Employed an ordinal Classification and Regression Trees (CART) model.
• Compared the performance of the ordinal CART model against conventional classifiers.

Main Results:

• The ordinal CART model demonstrated superior performance compared to conventional classifiers.
• The proposed methodology revealed patterns in absenteeism data not previously identified.
• Interpretability of the model facilitated a deeper understanding of employee absenteeism drivers.

Conclusions:

• Interpretable classification algorithms offer significant advantages for analyzing employee absenteeism.
• The developed ordinal classifier provides valuable insights for human resource management.
• This approach supports evidence-based decision-making to reduce employee absence and improve organizational outcomes.