Improving Financial Distress Prediction Using Financial Network-Based Information and GA-Based Gradient Boosting Method

Previous studies on financial distress prediction have chiefly used financial indicators which derived from financial statements as explanatory variables, so some potentially useful information that contained in the financial network was not considered. The listed companies can be represented as a complex financial network which the firms are regarded as nodes and the links account for stock returns correlation. The purpose of this study is to investigate whether network-based variables can improve the predictive power of financial distress prediction. Therefore, this study proposed a genetic algorithm (GA) approach to parameter selection in gradient boosting decision tree and integrated network-based variables for financial distress prediction. In order to verify the prediction capability of network-based variables and GA-based gradient boosting method in financial distress prediction, empirical study based on Chinese listed firms’ real data is employed, and comparative analysis is conducted. The experiment results indicate that the introduction of network-based variables and GA-based gradient boosting method for financial distress prediction can enhance predictive performance in terms of accuracy, recall, precision, F-score, type I error, and type II error.