Framework for the Development of an Enhanced Machine Learning Algorithm for Non-Cognitive Variables Influencing Students’ Performance using Feature Extraction

Abstract

Machine learning is a powerful tool for creating computational models in scientific analysis in areas where there is need to extract hidden data such as educational data. In order to make planning easier and identify at-risk students who may be in danger of failing or dropping out of school due to their academic performance, Educational Data Mining (EDM) uses computational tools. In this paper, a framework using machine learning approach was proposed to develop an enhanced algorithm for non-cognitive variables influencing students’ performance using feature extraction. In the framework, the Decision Tree (DT) and Linear Support  Vector Machine (SVM) are proposed as base classifiers, and Random Forest (RF) and Gradient Boosting (GB) as ensemble classifiers. The DT classifier allows the classification process to be modelled as a series of hierarchical decisions on the features, forming a tree-like structure. Using this technique, planning and predicting students who might be at-risk of dropping out would have been made easier.