Prediction of financial distress: Analyzing the industry performance in stock exchange market using data mining

The challenges and competition in the investment world recently became a great focus to be studied as it is greatly linked to profitability. It has been agreed that substantive negative earning of profitability of firms greatly linked to financial distress, a condition which a firm has difficulty fulfilling its financial obligations. Previous research developed financial distress prediction model using conventional statistical methods that suffer from disadvantages as it depends largely on some restrictive assumptions. This research used data mining methods as its superiority with less restrictive assumptions to predict financial distress, with both financial and non-financial variables examined. Focused in finance and infrastructure listed firms in Indonesia Stock Exchange (IDX) for 4 years period, logistic regression and C4.5 decision tree along with ensemble classifier are developed and evaluated. The decision tree with boosting model demonstrated the best performing prediction model, as it is in line with the previous research that boosting outperformed other methods, with the smallest error rates and highest accuracy, with overall accuracy, sensitivity and specificity are 94.61%, 94.6%, and 94.5% respectively. The result of this research also offer several inferences such as return on asset being the most significant or important predictive variable to determine financially distressed firms.