Pandemi Döneminin Türk Savunma Sektöründe Gönüllü İşten Ayrılma Nedenlerine Etkisi ve Ayrılma Niyetinin Tahminlenmesi

Abstract

In today's competitive environment employee turnover creates extra costs for organizations. In order to prevent this cost, the number of academic research has increased recently. Besides, with the growing use of analytical methods in the field of human resources, more studies have been conducted on employee turnover and prediction of employee turnover intention recently. This study is an exemplary application based on real data to predict the impact of the changing conditions during the pandemic period with Covid-19 on employees' turnover and turnover intention before and after the pandemic. Within the scope of the study, demographic, organizational and employment status data between 2017 and 2023 of the people working in the company which is operating in the defense sector in Turkey is used. In the analysis, supervised machine learning methods such as binary logistic regression, k nearest neighbors, support vector machine, decision tree and random forest are applied. In order to compare the models’ performances different approaches with different sample sizes are designed. Orange Data Mining and International Business Machines Corporation (IBM) IBM SPSS Statistics programs are used for modeling. As a result of the analysis, it was revealed that in the post-pandemic period organizational factors become effective in deciding whether to leave the job. In addition, it is found that the accuracy and AUC scores of the prediction models are approximately between 50 percent and 80 percent. Moreover, it was observed that model prediction performance is affected by many aspects such as the period to which the data belongs, the sample size, hyperparameter values and the number of features included in the model.