| Abstract [eng] |
130 pages, 20 pictures and charts, 42 references The aim of the master's thesis is to develop a bankruptcy forecasting model suitable for small and medium-sized enterprises. To achieve the aim, the tasks formulated in the work are: to study bankruptcy forecasting from a theoretical aspect using literature analysis; to compare bankruptcy forecasting models; to identify the features of bankruptcy forecasting in small and medium-sized enterprises; to develop a methodology for modeling bankruptcy forecasting in small and medium-sized enterprises; to conduct an empirical study of bankruptcy forecasting modeling in small and medium-sized enterprises. Thus, the work consists of three parts: a literature review, a description of the methodology and an empirical study using hybrid bankruptcy forecasting models. Methods used in the work. One of the methods applied to perform the tasks outlined is the analysis and synthesis of scientific literature and other sources of information. The work also applies descriptive and comparative methods necessary for comparing bankruptcy forecasting models. An empirical study was conducted to model bankruptcy forecasting in small and medium-sized enterprises. Descriptive statistical and correlation analyses were used during the study. Hybrid bankruptcy prediction models were also created. In order to evaluate the research results, data systematization, processing, analysis, and evaluation methods were applied in the work. Research conducted in the work. During the research, operating and bankrupt small and medium-sized enterprises were selected. Next, an analysis of the financial and non-financial indicators of the selected enterprises was performed, determining their influence on the risk of bankruptcy. The relative financial and non-financial indicators obtained during the financial analysis were used to adapt six machine learning bankruptcy prediction models. In this way, hybrid bankruptcy prediction models were created that are effectively suitable for small and medium-sized enterprises. Results obtained in the research work and main conclusions of the study. The analysis of non-financial indicators of operating and bankrupt small and medium-sized enterprises allowed us to present one of the main conclusions of this work: enterprises from different areas of activity, of different ages and operating in different regions differ significantly in certain financial indicators. Thus, it is appropriate to use not only financial but also non-financial indicators in bankruptcy prediction models. The second main conclusion of the study is that the logistic regression model combining financial analysis and machine learning is the most effective for small and medium-sized enterprises, with the highest AUC value (0,946). This conclusion is based on the evaluation and comparison of the results of hybrid bankruptcy prediction models. |
