Clustering and Profitability Forecast of Companies listed in Tehran Stock Exchange with the Decision Tree c5 Approach

The intense competition prevailing in the world today and investors should be more cautious about their decision given the prevailing conditions. But this information alone is not useful, so it is necessary to use data mining techniques to analyze and interpret data so that more informative information will be available to users. Therefore, the purpose of this study is to cluster and forecast the profitability of companies. For this purpose, Tehran Stock Exchange companies were considered as the statistical population of the research and 888 companies in the period of 1387-1395 were selected as the research sample. So, in the beginning after the initial preprocessing of the data, with Matlab and Clementine software, using SSE criteria and K-Means method, the companies were converted to 3 clusters and the result of these clustering were measured by the standard quality measures. Finally, by using the C5 decision tree, cluster analysis and variables affecting profitability were identified; so that from the 32 considered variables only 8 includes: Gross profit to total assets, sales to total assets, profit to equity, operating profit to net sales, accrued profit and loss to equity, net profit to net sales, total liabilities to total assets and current assets to total assets affect the profitability of companies. At last, by taking these variables into account, prediction of each cluster was done, and the accuracy of the predictions sequence was 86,34%, 88,15% and 68.81%