Application of Data Mining Techniques for Achievement of Learning Behaviors Prediction in Sukhothai Thammathirat Open University on STOU eLearning

Abstract

This research aims to utilize data mining techniques, specifically decision tree methods, to predict academic outcomes based on learning behaviors in STOU eLearning for Master's degree students at Sukhothai Thammathirat Open University. The sample group consisted of Master's degree students in the Nursing Science program who enrolled in the course set 51715, which includes Nursing Organization, Human Resources, and Quality Management, in the second semester of the 2022 academic year, totaling 39 students. The research methodology was divided into five steps: 1) collecting students' learning behavior and academic performance data from log files via STOU eLearning throughout one semester; 2) preparing the data; 3) reducing data dimensionality; 4) transforming the data; and
5) analyzing the data using the RapidMiner software. The research findings revealed that: 1) The model for predicting academic outcomes from the learning behaviors of Master's degree students is divided into three pathways: Pathway 1 includes students who attended no more than 376 times and for whom the program predicts a passing grade
(S or H). Pathway 2 includes students who attended more than 376 times and completed pre-test and post-tests, for whom the program predicts a passing grade (S or H). Pathway 3 includes students who attended more than 376 times but did not complete the pre-test and post-tests, for whom the program predicts a failing grade (U).  2) The effectiveness of predicting academic outcomes from the learning behaviors of Master's degree students showed that the accuracy of predicting passing was 87.88%, the accuracy of predicting failing was 50.00%, and the overall accuracy of predicting academic outcomes from learning behaviors was 84.62%.