Assessment of Enterprise Bankruptcy Probability

This paper is devoted to the development of an enterprise bankruptcy probability model. In accordance with the goals and objectives of the study consists of three main parts. The first part of the work includes theoretical aspects in the field of assessing the probability of bankruptcy. In the first chapter, the concept of “bankruptcy” was formulated and a classification was given of the existing types of bankruptcy prediction models. Also, an assessment was made of the frequency of use by researchers of each of the models described. The second chapter is devoted to the consideration of existing models for assessing the likelihood of bankruptcy, to analysis of their pros and cons, as well as the factors used in them. Much attention is paid to models built on the basis of Multiple discriminant analysis, and a model of neural networks belonging to artificial intelligence models is also considered. In conclusion of the second chapter, a summary histogram was created, which illustrates the frequency of use of explanatory variables in designing models for assessing the probability of bankruptcy of enterprises. The third part is the analytical part of the study, which includes collecting a database, analyzing the sample, building models, interpreting empirical results and comparing the models obtained. In this part, two models were built: a binary choice model related to the statistical type of models (Logit model) and the Random Forests model, which is an artificial intelligence model. Also in the final chapter of the study, a comparison is made of the constructed models based on the ROC curves and the AR coefficient. In the conclusion of this work lists the main results of the study, as well as formulated proposals for further research. The object of study is the probability of bankruptcy of enterprises. The subject of study are the features of assessing the probability of bankruptcy and the degree of influence of financial indicators on the bankruptcy of enterprises. The theoretical significance of this study includes comparing the binary choice model of the logit model and the Random Forest artificial intelligence model from the point of view of building bankruptcy prediction models. To the practical significance of the study can be attributed created in the framework of this work, an enterprise bankruptcy probability models. These models can be used by investors to assess the likelihood of bankruptcy of an enterprise, as well as by the enterprise itself to monitor the overall financial situation in the company and following correction its management.