Employees recruitment: A prescriptive analytics approach via machine learning and mathematical programming.

Area of Science:

• Data Science
• Human Resources Analytics
• Machine Learning

Background:

• Traditional recruitment methods often overlook crucial data insights, leading to suboptimal hiring and placement decisions.
• Existing HR analytics tools may lack interpretability, hindering their adoption and practical application by recruiters.
• Balancing recruitment success with diversity goals presents a significant challenge in organizational hiring.

Purpose of the Study:

• To develop a comprehensive analytics framework for HR recruiters to improve hiring and placement decisions.
• To introduce an interpretable machine learning model for predicting recruitment success at the individual placement level.
• To create a global recruitment optimization scheme considering multilevel organizational factors.

Main Methods:

• Application of Variable-Order Bayesian Network (VOBN) models to a large-scale, decade-long recruitment dataset.
• Development of a two-phase framework: local prediction of recruitment success and global recruitment optimization.
• Utilizing a uniquely large and heterogeneous dataset encompassing hundreds of thousands of employee records.

Main Results:

• The VOBN model achieved high accuracy and provided interpretable insights into recruitment data.
• The framework successfully predicted candidate placement success at the pre-hire stage.
• The devised framework demonstrated improved recruitment success rates and diversity compared to actual decisions, balancing the inherent trade-off.

Conclusions:

• Interpretable machine learning, specifically VOBN, offers valuable and often counter-intuitive insights for HR professionals.
• The proposed analytics framework serves as an effective decision support tool for enhancing real-world recruitment processes.
• The framework provides a balanced recruitment plan, optimizing both success rates and diversity in organizational hiring.