The instrumental machine learning methods for corporate bankruptcy prediction

Abstract

Subject. It is very important for corporate governance and a choice of partners to evaluate the company’s position. Therefore, bankruptcy forecast methods have been actively studied in theoretical and practical proceedings for a long time. Recurring crises and high market dynamics make the subject especially relevant. Objectives. I develop the instrumental method based on machine learning to predict corporate bankruptcy. The study also reviews data sources, the potential of forecasting models, and chooses inputs for company analysis. Methods. I applied methods of analysis and synthesis, and the systematization, formalization, comparative analysis. I referred to theoretical and methodological principles set forth in national and foreign proceedings on the company analysis and bankruptcy prediction. I investigate issues of data compilation, and building the artificial neural network for teaching the model. Results. I proposed and tested the instrumental method to predict bankruptcy. I suggest using my own sets of indicators for forecasting, which I selected by analyzing key indicators of financial sustainability, efficacy, and key external factors influencing market actors. The article presents a data sample for teaching purposes, which includes both the Russian and foreign companies, thus expanding its size. I devised machine learning models generating high-precision forecasts. Conclusions and Relevance. The findings contribute to bankruptcy prediction methods and can be used for administrative decision-making to automate their own analysis or analyze other entities, which the company cooperates with.