This paper introduces an intelligent tool with a novel CatML stacking classifier designed to enhance predictive analytics for postgraduate university admission chances. The proposed classifier uses the CatBoost algorithm as a core component of the stacking ensemble method, which integrates CatBoost and Multi-Layer Perceptron (MLP) learners to improve predictive performance. The dataset comprises 13 questionnaire-based surveys, including academic records, standardized test scores (i.e., GRE, IELTS/TOEFL), publication status, extracurricular activities, recommendation letters, and personal statements from Bangladeshi students who applied to various U.S. postgraduate programs. Experimental results demonstrate that the CatML stacking classifier outperforms conventional models, achieving superior accuracy (88.14%) and robustness in predicting admission outcomes. The enhanced performance is attributed to the model’s ability to capture complex, non-linear relationships within the data, facilitated by the CatBoost algorithm's handling of categorical features and prevention of overfitting. Finally, this model deploys in a web system developed with HTML, CSS, JavaScript and Flask. This research underscores the efficacy of advanced ensemble techniques in educational data mining and provides a valuable intelligent tool for students aiming to navigate the complexities of U.S. postgraduate admissions. The CatML stacking classifier offers significant improvements in predictive analytics, thereby assisting students in making informed application decisions.