Prediction of Student Academic Performance using Neural Network, Linear Regression and Support Vector Regression: A Case Study

Abstract

Predicting students" academic performance is very crucial especially for higher educational institutions. This paper designed an application to assist higher education institutions to predict their students" academic performance at an early stage before graduation and decrease students" dropout. The performance of the students was measured based on cumulative grade point average (CGPA) at semester eight. The students" course scores for core and non-core courses from the first semester to the sixth semester are used as predictor variables for predicting the final CGPA8 upon graduation using Neural Network (NN), Support Vector Regression(SVR), and Linear Regression (LR). The study has verified that data mining techniques can be used in predicting students" academic performance in higher educational institutions. All the experiments gave valid results and can be used to predict graduation CGPA. However, comparisons of the experiments were done to determine which approaches perform better than others. Generally, SVR and LR methods performed better than NN. Therefore, we recommend the adoption of SVR and LR methods to predict final CGPA8, and the models can also be used to implement Student Performance Prediction System(SPPS) in a university. Thus, the study has used the models from SVR and LR methods for designing an application to do the prediction task.

Prediction of Senior High School Students’ Performance in a State University: An Educational Data Mining Approach

“…It is noticeable in Table V that NB and NN have the lowest prediction error. Through counts, it can determine that NB only has seventeen (17) prediction error while NN has a total of eighteen (18) prediction error. Thus, it is now harder to decide which among the two classification algorithms is better since there is only 0.56% of difference in terms of prediction error between them.…”

“…Thus, unlike NB, LR assumes that the dependent variables are related to the independent variables or the class in a linear manner. Stated in the study of Bum et al [17] that LR is the quickest machine learning algorithm in developing model without compromising the accuracy of its prediction hence, it was proven on the experiment of research [18] after comparing the results of shortest time in building model and prediction accuracy to another two sophisticated machine learning algorithms (NN and support vector regressor). In this manner, LR can be used through this formula: y=B0+B1*x1 where y is the class and x are the independent variables.…”

Performance Evaluation of Machine Learning-Based Algorithms to Predict the Early Childhood Development Among Under Five Children in Bangladesh

“…Previous studies in Bangladesh observed that RF algorithms were the best predictors in terms of accuracy when predicting childhood anemia, while SVM identified some important characteristics when used to predict the status of child diarrhea in Bangladesh (Khan et al, 2019;Maniruzzaman and Abedin, 2020). In a different context, neural networks were poorer at predicting student academic performance than linear regression and support vector regression (Obsie and Adem, 2018). In another study, evidence from early childhood asthma persistence (Bose et al, 2021) and early childhood obesity (Pang et al, 2021) and researchers observed that XGBoost was found to be one of the best performing models in their study.…”

Feature Correlation with Student Education Performance

“…Belachew and Gobena's (2017) results were supported by the paper written by Jayaprakash, Balamurugan, and Chandar (2018). However, Obsie and Adem (2018) found that Linear Regression and Support Vector Regression were better than Neural Networks. Also, Acharya and Sinha (2014) found that the decision tree class of algorithms was the best.…”