Predicting student graduation outcomes is crucial for enhancing academic success rates and supporting at-risk students. This study developed a machine learning-based prediction system using Support Vector Machines (SVM), Random Forest, and Logistic Regression to classify students as likely to graduate on time. A synthetic dataset containing 4,424 instances and 35 features was utilised, encompassing demographic, socio-economic, and academic features. Data preprocessing included feature engineering, encoding, and scaling, ensuring the dataset was optimised for model training. Random Forest outperformed SVM (91%) and Logistic Regression (90%), achieving the highest accuracy at 92%. Results proved the robustness of ensemble methods, like Random Forest, in addressing complex datasets, whereas SVM demonstrated effectiveness in recall performance. The study underscores the utility of predictive analytics in academic contexts, offering actionable insights for early intervention and resource allocation. Future work should focus on validating the system with real-world datasets and exploring advanced algorithms to further improve accuracy and scalability. Keywords: Data Mining, Data Preprocessing, Ensemble Learning, Predictive Analytics, Supervised Learning Algorithms. Proceedings Citation Format Fatimah Adamu-Fika, Dawud Bala Madaki, Aanuoluwapo Enyojo Baba-Onoja, Aisha Tijjani Ramalan, Ahmed Taiye Mohammed, Kamaludeen Shehu Bature (2023): Modelled Machine Learning Algorithms to Predict Students Academic Performance in Tertiary Institutions. Proceedings of the 36th iSTEAMS Accra Bespoke Multidisciplinary Innovations Conference. University of Ghana/Academic City University, Accra, Ghana. 31st May – 2nd June, 2023. Pp 418-427 dx.doi.org/10.22624/AIMS/ACCRABESPOKE2023P39
Adamu-Fika et al. (Fri,) studied this question.