Explore the Contribution of Learning Style for Predicting Learning Achievement and Its Relationship with Reading Learning Behaviors

Abstract

Prediction is an important branch of research in learning analytics, in which the prediction of learning achievement has much practical value for improving instructional management and enhancing learning effectiveness. As a type of cognitive data, students’ learning style data offers great potential for predicting their learning achievement. Based on the analysis of the contribution of learning style data on prediction model creation, this study uses the Felder and Silverman learning style scale to examine 238 students’ learning styles as feature elements and explores the feature importance using six machine learning algorithms to create models for learning achievement prediction. Besides, to identify the relationship between learning styles and learning behaviors, and the hidden learning patterns behind learning styles, the study collected reading log data using the E-book system for correlation and principal component analysis. It was found that the Decision Tree model obtained the best results in terms of accuracy and other indicators. Secondly, the VisualScore feature showed the greatest influence on all the six models used. Thirdly, the study also found that learning styles were highly correlated with repeated learning and marking behavior in reading behavior. Finally, the analysis showed that the visual and verbal dimensions under the VisualScore features had three common learning patterns of repeated reading, marking, and mobile reading, in addition to differences in learning patterns in terms of time spent.