This study aimed to determine predictive models that would identify the most important predictor variables for students in the lowest proficiency group in public schools and private schools. After experimenting with different machine learning approaches, the random forest classifier (SVM) models were found to perform most accurately (Lundberg & Lee, 2017). To identify the most important predictors of being a poor performer in mathematics, Shapley values were generated, which produces a ranked list of several features in descending order. To manage complexity in comparing the key variables for private and public student performance classification, the 10 most significant features for the public and private school groups are analyzed and illustrated in Figure 1. Red bars represent direct relationships, whereas blue bars represent inverse relationships with identifying the poor-performing students in mathematics. Shapley Additive exPlanations (SHAP) values represent the level of variable importance relative to other variables. The bar length of each variable corresponding to the x-axis values shows the strength of the variable’s influence in the prediction of the model. The meanings of each important variable are summarized in Table 1, which also highlights the similar and contrasting results for private and public schools.