A Hybrid Approach Using Decision Tree and Multiple Linear Regression for Predicting Students’ Performance Based on Learning Progress and Behavior

Analyzing factors related to learning progress such as coursework scores, how many times students were occasion, plagiarism or failure, and time spent at the library helps to determine factors in the reduction of dropouts. Many researchers have used traditional methods to predict students' academic performance, and a few research studies have developed a new hybrid approach, a combined classification and prediction method in this field. This study has assessed students’ performance using a hybrid method including a decision tree and multiple linear regression to predict their possibility of graduation. Specifically, the decision tree model is used to classify the ‘Adequate’ and ‘Fair’ classes. Then, multiple linear regression models were used to predict future Cumulative Grade Point Average (CGPA). After evaluating the statistics, the first and second coursework scores exhibit a significant impact on the results. Other attributes such as time spent at the campus or the number of times that students failed in the previous semester should be considered in this context. The decision tree model’s accuracy is 0.47 and the Correlation Coefficient of the multiple linear models is 0.52. The result of this research is an equation with a specific weighted score toward the final results. This, in turn, would ensure early and appropriate actions from education to increase the academic achievement of such students.