Bankruptcy prediction is an important economic problem. It is a crucial problem in finance, as successful prediction enables stakeholders to take early actions to reduce their economic losses. Machine learning can effectively predict a company’s bankruptcy situation and provide timely reminders to the enterprise. In this study, a real-world dataset collected from the Taiwan market from 1999 to 2009 is used. The bankruptcy status is identified and set as the prediction target. An oversampling framework is used to balance the label distribution, and PCA is used for feature reduction. Four machine learning classifiers are implemented and compared, namely, support vector machine, random forest, AdaBoost and XGBoost. The numerical results demonstrate that the random forest classifier outperforms baselines in terms of the F1 score.