| dc.description.abstract | The low rate of timely graduation of students at the Bachelor level is a problem that university study programs often face every year. This has a negative impact on the accreditation assessment of the study program itself. Through this research, a study period reminder and graduation prediction system was developed using the Multilayer Perceptrons (MLPs) and Support Vector Machine (SVM) algorithms on the Telegram chatbot. The dataset used is academic data from undergraduate students in Computer Science and Information Technology at the University of Sumatera Utara semester 1 to semester 4, 5, and 6 to predict the graduation status of students in semester 5, 6, or 7 using the MLPs and SVM. However, the dataset turned out to have imbalanced classes, so to overcome this, Synthetic Minority Over-sampling Technique with Edited Nearest Neighbor (SMOTE-ENN) was used. The accuracy obtained was also relatively higher than without this technique, namely 98.28%, 99.41%, and 99.63% for student data for semesters 5, 6, and 7 respectively. Validation accuracy was 88.14%, 96.88%, and 94.12%. And test accuracy of 94.92%, 96.88%, and 98.53%. The implementation of the Telegram chatbot using this model to predict graduation was successfully carried out along with a reminder system by adding a feature to check students' cum laude status as measured by the criteria for grades D or E, as well as participation in competitions. | en_US |
