The scholarship selection process at a private university is still conducted manually by reviewing scholarship applicant documents individually, resulting in a time-consuming process and potential inconsistencies in evaluation. This study aims to develop a multi-class Machine Learning-based classification model to support scholarship classification based on recipient criteria and evaluating model performance using accuracy, precision, recall, F1-score, and confusion matrix metrics. This study applied the Cross-Industry Standard Process for Data Mining (CRISP-DM) framework consisting of six stages Business Understanding, Data Understanding, Data Preparation, Modelling, Evaluation, and Deployment. The dataset consists of 408 historical scholarship recipient records categorized into four scholarship classes. A key contribution of this study is the integration of the CRISP-DM methodology with feature importance analysis using Random Forest Feature Importances, class balancing using SMOTE, and machine learning classifiers, including Random Forest, SVM, and Naïve Bayes, within a unified predictive modelling framework. Based on the evaluation results, the Support Vector Machine algorithm achieved the best performance with an accuracy of 77% and a macro F1-score of 63%, followed by Random Forest at 76% and Naïve Bayes at 70%. The SVM model was then implemented as a Streamlit-based web to support scholarship recommendations and is not intended as a final scholarship approval system. This research contributes to the development of an efficient, data-driven scholarship classification support system for higher education institutions.