A Comparative Study of Accuracy for Novel Bankruptcy Prediction System Using Diverse Algorithms
摘要
This study evaluates the efficacy of multiple algorithms in forecasting corporate bankruptcies, using a dataset of 6,819 instances and 96 variables. We tested the Random Forest (RF), Multilayer Perceptron (MLP), Gradient Boosting (GB), and Support Vector Machine (SVM) algorithms for bankruptcy prediction. Classification accuracies were observed as 91.1% for RF, 95.8% for MLP, 98.2% for GB, and 94.2% for SVM. An independent sample t-test confirmed significant accuracy differences between the algorithms (p < 0.05), highlighting GB’s superior performance over RF, MLP, and SVM.