GRADUATION PREDICTION SYSTEM USING ARTIFICIAL NEURAL NETWORK

Educational Data Mining (EDM) is one area of research in the field of Data Mining which aims to explore knowledge or obtain models from an academic database. One of the benefits of EDM is to generate predictions for student academic achievement in the future. In this study, a Multi-Layer Perceptron (MLP) based EDM system is proposed to predict students' graduation at a university with five parameters: gender, college admission year, GPA semester 1, GPA semester 2 and GPA semester 3. Training data which used a total of 292 records with missing value in some of its features. The system test was done by Cross Validation method with three k-fold, ie 3, 5, and 10. The test result showed that MLP succeeded in classifying the data with a high level of accuracy of 94.8%, where the precision and recall values obtained 94.2% and 96.2% respectively. In addition, comparative tests were performed with several other classification methods, namely k-NN, SVM, Naive Bayes, and Decision Tree to find out how well the method proposed in this study was compared with other methods.