Task-based Learning Analytics Indicators Selection Using Naive Bayes Classifier And Regression Decision Trees

Abstract

The advent of the internet has strongly influenced the way we learn, by introducing e-learning systems as an aid to traditional education, sometimes even as the sole means of learning. An online learner can generate a multitude of learning analytics indicators that can be used to improve these learning systems using artificial intelligence algorithms. Nevertheless, the use of a large number of learning indicators causes overfitting that degrades the performance of machine learning algorithms. Therefore, in this paper, we will focus on the implementation of dynamic optimization of the number of learning indicators, based on the type of the considered task. This optimization will be done through two different machine learning algorithms: Naive Bayes Classifier for the classification tasks and Regression Decision Trees for the regression task. The adaptation of these two algorithms with various scenarios provides convincing results that demonstrate a significant improvement in the predictions made.