Návrh a implementace prediktivního modelu fluktuace zaměstnanců ve výrobním podniku s využitím vícerozměrných metod statistické analýzy

Abstract

Human Resource Management is one of the most important elements within organization's decision processing and represents strategic approach to the effective management of people in an organization. HRM decisions are becoming more frequently based on data which is the area of Human Resources Analytics. This thesis presents a quantitative approach of HRA to solve human resource management problem with employee turnover. Undesirable employee turnover always presented serious problem for both private and public organizations causing high financial costs, losses in know-how and decreased productivity. The aim of the thesis is to design and describe the construction of a predictive model of employee turnover in a production company using multivariate statistical analysis with a demonstration of practical application for a production company from Moravian-Silesian region. The thesis includes analysis of employee turnover on real data for the period from 2015 to 2019 and identifies and verifies factors which are significant for undesirable employee turnover. The individual factors are discussed and a predictive model is designed using multivariate logistic regression. There are two separate predictive models estimated in the thesis. The first model involves the entire population of employees to demonstrate and confirm the procedure for design, estimation and verification of a predictive model. The second model is estimated using selected group of qualified white-collar workers with potential of practical implementation in the company’s decision making. Estimated predictive model of undesirable employee turnover is used to calculate the individual risk of leaving for each employee and verified on out-of-sample data for the period of second half of 2019. The predictive model is assessed as a potentially important tool in HRM policy-making and a proactive approach to data-oriented solutions in the area of employee turnover and retention. The partial goals of the work are to map the mutual relations between the predictors of the created model using multiple correspondence analysis and position maps and a prescriptive model in the form of a strategic map capturing the complex process of model implementation. Finally, the possibilities of implementation and practical use of the predictive model and its strengths and weaknesses including opportunities and threats in the analysed organization are discussed.