Neural networks analysis in business failure prediction of Chinese importers: A between-countries approach

摘要

Past literature on business failure prediction rarely covered international discussions, and hardly any effort had been devoted to the establishment of a failure prediction model using a between-countries approach. In this study, we have proposed a means to collect and determine explanatory variables using a between-countries approach. In addition, we have established a systematic experiment to investigate the influences of techniques for both network architecture selection and variable selection on neural network models' learning and prediction capability. Another issue of the study is to explore the influences on different sample mixture ratios of the training and testing subsets, which relates to the stability and generalizability of models. Four assumptions of cost ratios of Type I and Type II errors (CI/CII ratios) are also examined in evaluating the model's effectiveness.The results show that the neural network models provide good classification capability in both cross-industry and industry-specific contexts. Moreover, the higher the training sample size and the larger the number of hidden nodes, the higher the classification rates, the lower the Type I error rates, the lower the relative CI/CII ratios. Among the three variables selection methods, factor analysis is superior to stepwise discriminant analysis (SDA) and ALL in terms of classification accuracy, generalization ability and error costs, while SDA provides the worst performance in all situations.