Machine Learning Approaches for Predicting Company Bankruptcy: A Comparative Study

Abstract The present study explores the utilization of machine learning classifiers for the purpose of forecasting firm bankruptcy. The dataset consisted of financial metrics and was used to evaluate six different classifiers which included; Support Vector Classifier, Logistic Regression, K-Nearest Neighbors, Naive Bayes, Decision Tree, and Random Forest. In terms of accuracy in the original data (96.77%) and scaled data (96.70%), Random Forest Classifier emerged as the best performing classifier. This research indicates that careful choice of a model is crucial and also implies that machine learning has a great potential in improving risk management and financial decision making. The implications of these result for various domains in finance suggest that hybrid models should be researched and explained in better detail by future work to further improve accuracy and transparency. Furthermore, the use of machine learning can raise predictive accuracy among financial institutions, which will lower risks thereby increasing overall performance that contributes to financial stability.