The study introduced and utilized Byesian Additive Regression Trees (BART) to classify the Cumulative Grade Point Average (CGPA) of students based on the perceived illnesses they experienced. This addressed the lack of application of BART method in classification predictive modeling. The data was collected from 292 respondents, including 10 illness predictors and the CGPA. First BART model showed that all predictors had higher values, indicating multicollinearity, which was later confirmed through Variance Inflation Factor (VIF) assessment. After the refinement of the model, only four predictors- stress, anxiety, headache, and stomachache- were retained. Despite adjustments, the classification BART model yielded modest accuracy at approximately 66%. Among CGPA categories, CGPA-C was the most accurately predicted, while CGPA-A showed no correct classifications. Partial dependence plots and posterior predictive checks confirmed that individual perceived illnesses had minimal predictive influence. These findings suggested that although health perceptions were commonly reported among students, they were not strong predictors of academic performance as measured by CGPA. The application of BART highlighted its ability, complexity, and flexibility for statistical predictive modeling in classification setting for consistently employing higher values to all variable importance plots which determined the genuine relationship between CGPA and perceived illnesses. Keywords: Bayesian Additive Regression Trees (BART), Cumulative Grade Point Average (CGPA), Perceived Illnesses, Classification Modeling, Academic Performance, Multicollinearity, Variance Inflation Factor (VIF), Predictive Analytics, Partial Dependence Plots, Student Health