Course performance prediction and evolutionary optimization for undergraduate engineering program towards admission strategic planning

Abstract

Admission systems around the world are different in characteristics and processes. In Thailand, five different admission rounds affect admission strategic planning, students, and universities. This research proposes the course performance prediction and optimization to predict student performance data and find the optimum criteria for recruiting students in each engineering major respected for the undergraduate engineering program's admission round. The research uses data from undergraduate students in Engineering Faculty in Thailand during 2018-2020. The data preparation methods, such as missing value handling, feature generation, and correlation analysis for each course, are used. Predictive analytics aims to predict three engineering courses' average course grades using the generalized linear model, deep learning, and gradient booted tree. The model is evaluated by using relative error, root mean square error, and absolute error. Gradient boosted tree outperforms the other algorithms, which are 0-0.4% relative error. Prescriptive analytics is consequently used to optimize factors to get the optimum students to the faculty and major by using evolutionary optimization algorithms. This model is used to optimize decision-making in admission strategic planning of Engineering Faculty by optimizing students' number in each major and admission round.