Genetic programming decision tree for bankruptcy prediction

Abstract

In this paper, we apply the CART, C5.0, GP decision tree classifiers and compares with logic model and ANN model for Taiwan listed electronic companies bankruptcy prediction. Results reveal that the GP decision tree can outperform all the classifiers either in overall percentage of correct or k-fold cross validation test in out sample. That is to say, GP decision tree model have the highest accuracy and lowest expected misclassification costs. It can provide an efficient alternative to discriminates financial distress problems in Taiwan.