Optimizing human resources’ selection criteria and classification algorithms’ parameters in Greek public sector. A meta-analysis

Background: Successful human resources selection is considered the main step for every organization. Previous research has identified many challenges and innovations concerning the application of Artificial Intelligence/Machine Learning in Human Resources Management. Purpose: The purpose of this study was to apply machine learning algorithms in order to match human qualifications to position’s standards and finally to establish a rapid and more reliable procedure either for initial selection or for authority positions and additionally, to optimize the selection coefficients of properly chosen variables that describe qualifications and, in parallel, to optimize the best fit algorithms’ parameters in order to achieve the greatest accuracy. Finally, this procedure may support automatic mobility. Approach: This study was based on civil section data in order to match human qualifications to position’s standards using machine learning algorithms and finally to establish a rapid and more reliable procedure mainly for initial selection but also for authority positions. Supervised machine learning algorithms were applied. Optimization of selection coefficients of properly chosen variables was performed, followed by algorithms’ parameters optimization in order to achieve the greatest accuracy. Findings: Metrics of algorithms were improved at about 3% for accuracy and F- Measure, especially for J48, which found to be the best algorithm for matching with accuracy close to 97% and pruning simplified the final tree and thus visual classification. This procedure may also be useful in order to support a system of automatic mobility (internal and external) of highly qualified executives