Performance evaluation for four types of machine learning algorithms using educational open data

Abstract

Based on the educational data published under Creative Commons License, this study describes about the performance prediction experiment applied with four types of machine learning algorithms, including the deep learning algorithm, and examines how the prediction accuracy is affected depending on the selected feature quantities. The aim of this paper is to compare method selection and feature selection in terms of their ability to improve the prediction results. In data analysis by machine learning or deep learning, the determinant of result is often unclear. In the field of learning analytics, analysis can be performed even if the amount of data is small compared to the field of image recognition. Therefore, it is meaningful to compare analysis accuracy using machine learning and deep learning and to examine which method is most effective for prediction academic performance. In this research, we revealed that Deep Learning has the best method for Learning Analytics. Also, the results of this study indicate that feature selection is more important for improvement to prediction rather than method selection.