Feature selection methods application towards a new dataset based on online student activities / Muhammad Hareez Mohd Zaki ... [et al.]

The increasing usage of classification algorithms has encouraged researchers to explore many topics including academic-related topics. In addition, the availability of data from various academic information management systems in recent years had been increasing, causing classification to become a technique that is in demand by the educational institutes. Thereby, having a classification technique is important in researching the data of students’ performance. The problem during the classification of students’ performance is the lack of factors used to identify and evaluate their performance. Most of the articles used students’ grades as the most influential factor to predict students’ performance. Thus, more significant features are needed to evaluate students’ performance to improve the existing method. Due to the reason, a dataset is proposed to introduce some features that can affect the students’ performances. The dataset's features are based on online students’ activities during e-learning. This study will perform Analysis of Variance Test (ANOVA), Chisquared Test, Recursive Feature Elimination (RFE) and Extra Tree algorithm (ET) as feature selection methods to pre-process the proposed dataset that is considered raw data. The experimental results showed that ‘Answered all questions’, ’Afterclass notes’, ‘Correct 3 and above’ and ‘In-class notes’ were the most significant features in evaluating students’ performance. The study is significant towards educational data mining in analysing the students’ performance during online students’ activities.