Predicting Indonesian Retail and Manufactures Companies Bankruptcy Using Machine Learning

Corporate bankruptcy prediction tools certainly have diverse roles in finance, especially in the retail and manufacturing sectors that are experiencing economic shifts in the country with various challenges. Conventional approaches such as Altman's Z-Score method and Zmijewski Z's model have limitations in handling the complexity of existing financial relationships, thus encouraging the utilization of machine learning techniques. This research analyzes the accuracy of 15 machine learning systems and algorithms consisting of: Logistic Regression, SVM Linear Kernel, K-Neighbors Classifier, Ridge Classifier, Ada Boost Classifier, Linear Discriminant Analysis, Extra Trees Classifier, Random Forest Classifier, Gradient Boosting Classifier, Decision Tree Classifier, Light Gradient Boosting Machine, Dummy Classifier, Quadratic Discriminant Analysis, Extreme Gradient Boosting, and Naive Bayes. Implementation is carried out to predict company bankruptcy using financial data on the Indonesia Stock Exchange IDX for the period 2020 to 2023. Research shows that Logistic Regression is the best performing model compared to other models, with accuracy of 90 percent, recall of 100 percent, and F1 score of 91.67. Analysis of the importance of model features emphasizes that total assets are the most dominant factor in predicting bankruptcy. These findings further strengthen the role of machine learning in supporting financial risk evaluation and provide strategic value to financial decision makers.