The Comparison of Data Classification Efficiency to Predict Scholarship Awarded for Undergraduate Students by Data Mining Techniques

Abstract

This research analyzed factors involved with the scholarship data of undergraduate students and compared the performance of data mining techniques used to analyze the data: Decision Tree, Naïve Bayes, K-Nearest Neighbors, Deep Learning and Random Forest. The data was collected from the scholarship form of undergraduate students in the Faculty of Business Administration and the Faculty of Art from Borpitpimuk Chakrawad Division, Rajamangala University of Technology Rattanakosin during 2014 to 2018 school years. The dataset had 15 attributes and 1,155 records. The RapidMiner Studio program was used to find the number of relations attributes and model, and the 10-Fold Cross Validation to evaluate the model. The results showed that 11 attributes were involved. The accuracy of Random Forest, Deep Learning, Decision Tree, Naïve Bayes and K-Nearest Neighbors were at 94.28%, 93.51%, 92.64%, 92.47%, and 89.70%, respectively. The research could be used to predict, to analyze recommendations and to develop an information system to support the decisions in awarding scholarships to students.