İşe alım öncecinde adayların gelecek başarılarının makine öğrenmesiyle tahmini: Bankacılık sektöründe bir vaka çalışması

With the developing competitive environment and conditions, it has become important for companies to employ successful employees to survive. To create a strong workforce profile, companies first need to employ the right personnel. For this reason, the recruitment process is seen among the Human Resources (HR) processes with the highest return on investment (ROI). The rapid digitalization brought by the Covid-19 pandemic process and the compliance requirements of this process have made new business models necessary in the recruitment processes. Companies are now advancing their recruitment processes through online platforms in parallel with their current processes in recruiting new personnel. In addition to this, the increase in job applications and the increasing number of position requirements have increased the complexity of the process and the candidate data that need to be evaluated. Wrong hiring decisions caused by the difficulty of evaluation and complexity can cause long-term and financial and non-financial losses for companies. For this reason, companies need more data-driven decision support systems to tackle these challenges. By analyzing complex data with machine learning (ML) approaches, companies can offer meaningful outputs to decision-makers in their recruitment processes. Most of the analytical studies on recruitment processes focus on modeling the decisions of decision-makers. However, these models are sensitive to the subjective and biased evaluations of decision-makers. This study, on the other hand, focused on predicting the future performance of new candidates by learning the historical data of the employees who were evaluated as successful and unsuccessful in the recruitment process with machine learning algorithms. In this way, it is thought that decision-makers will be supported to employ the right employee by reducing the difficulty and complexity of evaluating the increasing data in the recruitment process. This study covers the data of 597 employees of a private bank serving in Turkey. In the creation of successful and unsuccessful output labels, the performance evaluations of the employees in the first two years have been taken into account. A three-stage methodology has been followed in the study. The first step of this study is to obtain and prepare the data set to be included in ML. In the second stage, the prepared data set has been divided into training and testing. Then, five-fold validation has been performed for Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNNs), Decision Trees (DTs), and Multi-Layer Perceptron (MLP) algorithms with the training data. According to the calculated evaluation criteria, a Logistic Regression model with an accuracy of 71.19% has been proposed. In the last stage, the prediction performance has been developed by optimizing the parameters for the proposed model. With the best parameter values of the developed model, a 73.14% accuracy rate has been obtained in the training data. Then, the model has been run with a test data set that it had not seen before, and a successful accuracy rate of 71.67% has been achieved.