Hybrid Machine learning algorithms at the service of student performance

Abstract

Abstract The ability to alter and improve a student's status in order to get the greatest performance out of them so that they pass their courses is an important component of today's educational landscape. This operation allows for the prediction of a student's performance in one or more disciplines. This has become possible nowadays through the use of Machine Learning algorithms that mine educational data to predict student performance by training the models and testing them with the available data set while using different algorithms. In this study, we compared 9 algorithms namely: ANN (Artificial neural network), DT (Decision tree), ELM based Model, KNN (K-nearest neighbour), LR (Logistic regression), LR 1 (Linear regression), NB (Naïve Bayes), RF (Random forest) and finally SVM (Support vector machine) in order to obtain the best models based on students’ performance in well-defined disciplines in order to improve their results and the success in their study. We started with the data collection and then we carried out a preprocessing process, after which, we built models in order to compare and evaluate them. Subsequently, we compared the results obtained and which were generated by the different algorithms in table 2 and which showed that the Random forest had the best ranking and this, in almost all the methods used (MAE , RMSE, ACC, AUC, RAE CSR) monitored by SVM which had satisfactory results.