Information Content of Intellectual Capital and Financial Performance Indicators in Financial Prediction by Data Mining Approach

Financial distress prediction is always a concern for shareholders, creditors and business unit managers. Hence, various models proposed to predict financial distress. In this research, decision tree model and linear multiple discriminant analysis model have been used to predict financial distress of companies. In addition, the results of these two models are compared with each other. In this regard, it has been tried to use intellectual capital and value-based performance in addition to conventional financial ratios in predicting financial distress. To achieve the objectives, three hypotheses formulated. The purpose of the research hypotheses is to investigate the information content of intellectual capital and performance indicators and financial ratios in predicting financial distress in Tehran Stock Exchange companies using decision tree models, linear multiple discriminant analysis and the combined model derived from factor analysis - linear multiple discriminant analysis and the factor analysis is the decision tree. To test these hypotheses, a sample selected from Tehran Stock Exchange companies during the years 2010-2010. The results indicate that the prediction of financial distress is possible using a decision tree, but in the decision tree model, intellectual and functional capital indicators do not have information content to predict financial distress. The linear multiple discriminant analysis model has been successful in combining financial ratios and performance indicators in the prediction of financial distress. Comparison of the capabilities of the two models indicated that although the accuracy of predictive analysis of the linear multiple discriminant analysis based on the assessment criteria (competing table) and the level below the rock curve (0.920) was higher than the decision tree (0.901), but based on the t-test, This difference is not significant (p = 0.207). In other words, the results of the two methods are very close.