Bankruptcy prediction using multivariate grey prediction models

Purpose In the face of complex and challenging economic and business environments, developing and implementing approaches to predict bankruptcy has become important for firms. Bankruptcy prediction can be regarded as a grey system problem because while factors such as the liquidity, solvency and profitability of a firm influence whether it goes bankrupt, the precise manner in which these factors influence the discrimination between failed and non-failed firms is uncertain. In view of the applicability of multivariate grey prediction models (MGPMs), this paper aimed to develop a grey bankruptcy prediction model (GBPM) based on the GM (1,N) (BP-GM (1,N)). Design/methodology/approach As the traditional GM (1,N) is designed for time series forecasting, it is better to find an appropriate permutation of firms in the financial data as if the resulting sequences are time series. To solve this challenging problem, this paper proposes GBPMs by integrating genetic algorithms (GAs) into the GM (1,N). Findings Experimental results obtained for the financial data of Taiwanese firms in the information technology industries demonstrated that the proposed BP-GM (1,N) performs well. Practical implications Among artificial intelligence (AI)-based techniques, GBPMs are capable of explaining which of the financial ratios has a stronger impact on bankruptcy prediction by driving coefficients. Originality/value Applying MGPMs to a problem without relation to time series is challenging. This paper focused on bankruptcy prediction, a crucial issue in financial decision-making for businesses, and proposed several GBPMs.