Imbalanced Data Classification Modelling Using CTGAN and Decision Tree for Student Graduation Predicting in a Courses

Abstract

Student graduation in one course is the main factor in supporting the learning process and minimizing the occurrence of drop outs. In this case, a prediction model is needed to be able to identify student graduation at the beginning of the learning process. The aim of this research is to produce a prediction model that has significant accuracy in predicting student graduation in one course using the decision tree algorithm and implementing the conditional tabular generative adversarial networks (CTGAN) model. CTGAN is a model that can produce synthetic data on certain input variables. First, the graduation dataset is collected and pre-processed, then a labeling process is carried out on the dataset, so that the data can be used as initial input for CTGAN modeling. Next, the dataset with certain label features is subjected to an oversampling process using the CTGAN model. Finally, a prediction process was carried out using a decision tree algorithm to produce significant accuracy values by utilizing the results of data processing with the CTGAN model. The results show that the decision tree algorithm has a very significant accuracy value in predicting student graduation performance using the CTGAN model processing dataset, as well as a significant precision value in predicting student failure in the learning process of a course.