Design and Development of the Students Performance Prediction Modul Using Open Education Resource

Abstract

Data mining is a tool to assist in decision-making. It is a modern technique, and it is still at a low level of implementation in this approach in Malaysia. In the market sector, data mining methods are usually used specifically to understand, predict, and schedule, which will improve organizational results. This project aims to predict student performance in MyGuru using activity log data. The prediction in this study helps to see a clear vision of student patterns and the activities they access most when logging in to MyGuru. With the resulting prediction model, student performance can be detected, which is influenced by the activities while accessing MyGuru. In developing this research, the main important part is extracting features. Feature extraction needs to be done clearly so that accuracy values ​can be achieved. The features in this study are selected from the attributes in the activity log. After converting the raw data, the data becomes a new dataset that is used to create a model according to the classifier. The classifiers used in this research are Random Forest, Support Vector Machine, Naive Bayes, K-Nearest Neighbors, and J48, which are involved in developing the model. The most accurate classifier for predicting student performance was Random Forest, which was 96.9%. Students, while using the MyGuru system, with the total of courses viewed was 15.68%, out of a total number of 644 people.The findings show that accuracy cannot be obtained if the original dataset has some issues, such as unbalanced data. Imbalanced data can affect accuracy.