Comparison and Prospect of Two Heaven Approaches: SVM and ANN for Identifying Students' Learning Performance

Education through AI has become a consensus in the educational technology community, in which learning performance prediction and classification are two popular tasks for educational data mining. Researchers investigated into different methodologies of machine learning according to their tasks. Often used methodologies are "Linear Regression, Neural Networks, Support Vector Machines, Naïve Bayes Classifier, and Decision Trees". Our analysis in this paper has shown that Support Vector Machines are less used than Neural Networks in the area of educational data mining. Although Support Vector Machines perform better in some investigations, there is a limited usage of it. We discuss several theses to find reasons for the occurring gap. Therefore, we focus on handling, Neural Networks as a trend topic, data and pre-processing, complexity of modeling, training and prediction speed, and problems of overfitting. Findings show that Support Vector Machines can be used by applying an acceptable effort. Their advantages were discussed regarding the challenges of educational technology.