Empirical implementation of extraction of decision-maker’s evaluation method in job candidates selection based on interpretable data mining tools

This paper presents the development of a Decision Support System (DSS) designed to assist human resource (HR) managers in selecting qualified job candidates from large pools of resumes. The system is built using data-driven methods, requiring no direct input from managers for initial decision-making process (DMP). It extracts important criteria by analyzing past managerial decisions and leverages interpretable machine learning techniques. The core ranking mechanism is based on the Choquet Integral (CI), which enables the system to account for interactions between different criteria, resulting in rankings that closely align with human decision-making patterns. A supervised learning method, decision tree (DTM) algorithm, is used to determine the criteria weights with the helping of, while K-Means clustering helps identify representative candidates in both accepted and rejected groups. The DSS was validated through multiple levels of testing, including the comparison of system outputs with managers' preferences and the evaluation of eight HR managers in a real-world setting as well as comparison with other methods. The results showed a strong correlation between the system's predictions and the actual decisions made by the managers. Candidates proposed by the DSS were generally perceived as better matches than those managers would have chosen manually. This finding demonstrates the system’s ability to prevent the oversight of qualified candidates due to the overwhelming volume of resumes.