Model selection for financial statement analysis: Comparison of models developed by using data mining technique

The purpose of this study is to clarify the applicability of data-driven approach in accounting area. As the first stage, focusing on the model comparison, this paper shows the effectiveness of model selection with data mining technique for the development of earnings prediction model based on financial statement data. In accounting area, researchers have not considered the characteristic of financial statement data and a suitable method for that enough. In this study, several methods are applied to develop prediction models. From the comparison with obtained results, we discuss the characteristic of financial statement data and show the possibility of data mining technique for the improvement of prediction accuracy.