Corporate Financial Distress Prediction Based on Ensemble Learning

Abstract

In order to decrease the uncertainty and instability of single classifiers in corporate financial distress prediction, this paper proposes a prediction model based on ensemble learning. The proposed approach first establishes different predictor systems by randomly partitioning dataset and applying feature selection techniques, and then constructs different classifiers based on different predictor systems. At last, these classifiers are combined for corporate financial distress prediction. In the empirical study, logistic regression and support vector machine are employed as the basic classifiers. The experimental results on 300 corporations listed in Shanghai and Shenzhen Stock Exchange show the accuracy and advantage of the proposed prediction model.