Predicting business failure using classification and regression tree: An empirical comparison with popular classical statistical methods and top classification mining methods

摘要

Predicting business failure is a very critical task for government officials, stock holders, managers, employees, investors and researchers, especially in nowadays competitive economic environment. Several top 10 data mining methods have become very popular alternatives in business failure prediction (BFP), e.g., support vector machine and k nearest neighbor. In comparison with the other classification mining methods, advantages of classification and regression tree (CART) methods include: simplicity of results, easy implementation, nonlinear estimation, being non-parametric, accuracy and stable. However, there are seldom researches in the area of BFP that witness the applicability of CART, another method among the top 10 algorithms in data mining. The aim of this research is to explore the performance of BFP using the commonly discussed data mining technique of CART. To demonstrate the effectiveness of BFP using CART, business failure predicting tasks were performed on the data set collected from companies listed in the Shanghai Stock Exchange and Shenzhen Stock Exchange. Thirty times’ hold-out method was employed as the assessment, and the two commonly used methods in the top 10 data mining algorithms, i.e., support vector machine and k nearest neighbor, and the two baseline benchmark methods from statistic area, i.e., multiple discriminant analysis (MDA) and logistics regression, were employed as comparative methods. For comparative methods, stepwise method of MDA was employed to select optimal feature subset. Empirical results indicated that the optimal algorithm of CART outperforms all the comparative methods in terms of predictive performance and significance test in short-term BFP of Chinese listed companies.