An Ensemble Stacking Algorithm to Increase Bankruptcy Prediction Model Accuracy

Abstract

To predict bankruptcy, bankruptcy analysis is necessary. Predicting bankruptcy incorrectly
frequently results in bankruptcy. High-accuracy machine learning for reversal analysis has to keep getting better.
To forecast bankruptcy, numerous machine learning algorithms have been used. To increase forecast accuracy,
model improvisation is still required. We suggest a combination model based on the stacking ensemble approach
and genetic algorithm-support vector machine (GA-SVM) to increase the accuracy of bankruptcy prediction. The
Taiwan Economic Journal's Taiwanese Bankruptcy dataset is used in this study. Next, in order to deal with
unbalanced datasets, we employ a synthetic minority over-sampling strategy. We use GA-SVM to choose the
best feature, stack the classifier as a novel approach, and employ extreme gradient boosting as a meta-learner.
With an accuracy of 99.58%, the results demonstrate the better precision achieved by the GA-SVM based on the
stacking model. Compared to using a single classifier, the accuracy achieved is higher. This study demonstrates
that the suggested approach has a higher accuracy rate in predicting bankruptcy