Student post-graduate success plays an essential role for higher education institutions (HEIs) to improve their standing and reputation, increasing their enrollment rates.  As a result, HEIs are allocating significant effort and resources to cultivate student success after graduation. In this study, student post-graduate success is categorized into four distinct levels including "Failure", "Satisfaction", "Success", and "High Success". To predict these outcomes, we employed four widely recognized machine learning methods: Random Forest Classifier (RFC), Decision Tree Classifier (DTC), Logistic Regression Classifier (LRC), and Support Vector Machine Classifier (SVMC). Our dataset comprised 373 student records compiled from NIPTICT/CADT, encompassing six generations across the majors of Computer Science, Telecommunications and Networking and Digital Business. The data were organized into three distinct categories: student profile, educational model, and student success. To prepare this raw data for training and testing our machine learning algorithms, manual annotation was performed. The weight score for each student was calculated based on our proposed post-graduate success metric. This calculated weight score is then used to classify students into their respective success levels. The result showed that the DTC achieved the highest performance, outperforming other classifiers with an accuracy of 0.96 on the validation set and 0.98 on the test set. RFC followed, demonstrating an accuracy of 0.92 on the validation set and 0.93 on the test set. Conversely, LRC and SVMC showed comparatively lower performance, achieving accuracies of 0.88 and 0.8 on the validation set and 0.89 and 0.84 on the test set respectively. The unexpectedly high performance of the Decision Tree prompted deeper investigation. We performed k-fold cross-validation to both validate these results and explore the factors contributing to its surprising effectiveness. Through cross-validation, DTC exhibits a slightly higher mean and higher standard deviation than RFC. This substantially higher standard deviation suggests that DTC's performance is less consistent and potentially less reliable when classifying student post-graduate success in real-world application. When these models are deployed and retrained with new data, the Decision Tree could exhibit greater fluctuations in predictive accuracy or other performance metrics. In contrast, Random Forest typically mitigates individual tree biases and variances, leading to more stable overall performance. Consequently, the Random Forest model is more suitable for this dataset and student post-graduation success prediction.