Educational Innovation: Predicting Graduation and Study Period Reminders Using the Multilayer Perceptrons and Support Vector Machines Algorithm on Telegram

The issue of low on-time graduation rates at the bachelor's level is a recurring problem for university study programs, negatively affecting their accreditation assessments. This research developed a system to remind students of their study period and predict graduation using Multilayer Perceptrons (MLPs) and Support Vector Machine (SVM) algorithms within a Telegram chatbot. The dataset comprised academic data from Computer Science and Information Technology students at the Universitas Sumatera Utara, covering semesters 1 to 4, and were used to predict graduation status in semesters 5, 6, or 7. Due to class imbalances in the dataset, the Synthetic Minority Over-sampling Technique with Edited Nearest Neighbor (SMOTE-ENN) were applied, resulting in improved accuracy rates of 98.28%, 99.41%, and 99.63% for semesters 5, 6, and 7 respectively. The validation accuracy is 88.14%, 96.88%, and 94.12% respectively, while test accuracy is 94.92%, 96.88%, and 98.53% respectively. The Telegram chatbot were successfully implemented to predict graduation and included a reminder system with a feature to check students cum laude status based on grades D or E and participation in competitions.