Implementation of Data Mining for Drop-Out Prediction using Random Forest Method

Abstract

Accreditation is one of the quality measurements for a University. Some elements of these measurements are students and graduate students. Prevention of students to drop out is a problem that is considered very important for the university itself. High levels of drop out students will have a bad impact on the university, such as bad reputation or low-grade accreditation. This research presenting the results of a case study analysis in educational data, by analyzing the data using the data mining technique. The author using the classification method, that focuses on drop-out prediction of undergraduate and diploma students at the ABC Faculty at XYZ University. To predict drop-out classification, academic data are needed. The raw data are student's academic data that enroll in university from 2008 to 2012. The raw data preprocessing then carried out to handle imbalanced data. This research uses synthetic minority oversampling technique (SMOTE) to handle imbalance dataset and random forest algorithm to predict drop-out within 2492 data. As a research result, the random forest algorithm accompanied by SMOTE can provide the best accuracy results by 93.43%. The main results of this research can be used to reduce drop-out levels by predicting potential drop out students and identifying potential factors related to drop out students.