Prediction of the Future Success of Candidates Before Recruitment with Machine Learning: A Case Study in the Banking Sector


DEMİRCAN M. L., Aksac K.

4th International Conference on Intelligent and Fuzzy Systems (INFUS), Bornova, Turkey, 19 - 21 July 2022, vol.505, pp.24-35 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 505
  • Doi Number: 10.1007/978-3-031-09176-6_3
  • City: Bornova
  • Country: Turkey
  • Page Numbers: pp.24-35
  • Keywords: Decision making, Recruitment process, Machine learning
  • Yıldız Technical University Affiliated: No

Abstract

Companies need to employ successful employees to survive in today's competitive and technology-oriented business environment. Most companies started to convert their recruitment processes to online platforms to face digital challenges and reach a massive number of candidates. In addition, the increase in job applications and the number of positions has raised the evaluation process complexity of candidate data. Wrong hiring decisions caused by the difficulty and complexity of evaluation can cause financial and non-financial losses for companies. For this reason, companies need more data-driven and machine learning (ML) based decision support systems to tackle these challenges. This study focuses on predicting candidates' future performance based on the historical data of successful and unsuccessful employees using ML. We aim to support HR decision-makers to employ the right employee by minimizing the evaluation complexity of big data in the recruitment process. This study covers the data of 597 employees of a private bank serving in Turkey and considers the first two-year performance evaluations of the employees while creating output labels. We have conducted a three-stage methodology: We prepared the data set and organized it as training and testing in the first two stages. Finally, we selected a Logistic Regression model with a 71.19% accuracy by performing five-fold validation for Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNNs), Decision Trees (DTs), and Multi-Layer Perceptron (MLP) algorithms. The result is improved by optimizing the parameters to a 73.14% accuracy level. The developed model is tested with another fresh data set and recorded a 71.67% accuracy rate.