Prediction Model based on Continuous Data for Student Performance using Principal Component Analysis and Support Vector Machine

Predicting student performance in higher education based on students’ self-efficacy and learning behaviour data is challenging, because the data is changing with time. The potential of using continuous data which is collected weekly needs to be investigated to identify the effectiveness in making predictions of low-performing students. Therefore, this paper presents the analysis of continuous data using the Principal Component Analysis (PCA) and Support Vector Machine (SVM) for predicting student performance. Firstly, we proposed three patterns of the Principal Component (PC) scores to predict the trends of behaviour within a semester. Secondly, we present an analysis of using different combinations of time frames in predicting the performance using the SVM. The obtained results show that three behaviour patterns have been extracted from the Hotelling’s T² values calculated using the PC scores which were fluctuating, ascending, and descending. The use of different time frames using SVM shows different accuracy results in prediction. The use of continuous data indicates that certain data can be predicted at the early stage using multiple time frames.