<![CDATA[Company bankruptcy becomes a serious problem because it can cause economic damage and other social consequences. It’s very important to predict bankruptcy as early as possible because prediction can be useful for evaluation and planning to avoid bankruptcy. Bankruptcy prediction is one of the imbalanced classification problems because the data with the bankrupt class is far less than the non-bankrupt class. This study aims to produce a good classification model for predicting bankruptcy. Resampling used a combination of SMOTE and under sampling, is applied to the training data to produce more optimal classification model. The classification method used for prediction is multilayer perceptron and complement naïve bayes. Predictive performance was calculated using recall, ROC AUC, and PR AUC. Based on the test, using SMOTE and under sampling is quite significant in improving the classification model on the multilayer perceptron. Resampling in complement naïve bayes also increased. recall and PR AUC scores The best recall obtained was 95.45% with the complement naïve bayes method. The highest ROC AUC with resampling was also obtained using complement naïve bayes of 87.80%. Therefore, it’s concluded that bankruptcy prediction using resampling with SMOTE and under sampling, can produce good performance for detecting bankruptcy.]]>
Copyrights © 2021
