Student Performance Prediction Model for Predicting Academic Achievement of High School Students

Abstract

Modern technology is necessary and important for improving the quality of education. While machine learning algorithms to support students remain limited. Thus, it is necessary to inspire educational scholars and educational technologists. This research therefore has three main targets: to educate the holistic context of rural education management, to study the relationship of continuing education at the upper secondary level, and to construct an appropriate education program prediction model for high school students in a rural school. The data for research is the academic achievement data of 1,859 students from Manchasuksa School at Mancha Khiri District, Khon Kaen Province, Thailand, during the academic year 2015-2020. Research tools are separated into 2 sections. The first section is a basic statistical analysis step, it composes of frequency analysis, percentage analysis, mean analysis, and standard deviation analysis. Another section is the data mining analysis phase, which consists of discretization technique, XGBoost classification technique (Decision Tree, Gradient Boosted Trees, and Random Forest), confusion matrix performance analysis, and cross-validation performance analysis. At the end, the research results found that the reasonable distribution level of student achievement consisted of four clusters classified by academic achievement. All four clusters were modeled on predicting academic achievement for the next generation of students. In addition, there are four success models in this research. For future research, the researcher aims to develop an application to facilitate instruction for learners by integrating prediction models into the mobile application to promote the utilization of modern technology.