Data Mining in Predicting Firms Failure: A Comparative Study Using Artificial Neural Networks and Classification and Regression Tree
Norashikin Nasaruddin, Wan-Siti-Esah Che-Hussain, Asmahani Nayan, and Abd-Razak Ahmad
Abstract. Financial Institutions and investors alike are very much interested in the accuracy of predicting the potential failures of firms. These financial institutions believe accurate prediction will lead to a low default rate in servicing their financial loans. The aim of this study is to find a better model to classify firms that is more likely to fail. Bad prediction model will lead to a high default rate. Using financial and non-financial information, this paper illustrates the construction and comparison of two models - artificial neural networks (NN) and classification and regression tree (CART) models to classify the failed from the non-failed firms. This study found that based on the training sample’s result (NN = 94.03% & CART = 94.69%) the overall accuracy result of CART is higher than the NN model. Similar result can be drawn for the validation sample with CART leading at 92.93% overall accuracy rate compared to NN’s 91.92%.