A Prediction Framework of Learning Outcomes Based on Meaningful Learning Features

• Published in: 2020 Ninth International Conference of Educational Innovation through Technology (EITT) pp. 205 - 210
• Main Authors: Tian, Hao, Lai, Song, Wu, Fati
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2020
• Subjects: Clustering algorithms; Electronic learning; feature engineering; Feature extraction; learning analytics; learning outcomes prediction; learning pattern; Prediction algorithms; Social networking (online); Support vector machines
• ISSN: 2166-0549
• DOI: 10.1109/EITT50754.2020.00043

How to use meaningful learning features to predict learning outcomes is a key issue in learning analytics. With feature engineering, this study constructed a feature set, including eight effective learning features about interactions of student-student, student-teacher, student-content, and student- interface. The trace data of 108 middle school students on the E- cloudbag LMS was collected and analyzed, which proved that the feature set can effectively predict the students' learning outcomes. The random forest algorithm achieved the best prediction effect, and the prediction accuracy rate was 73.15%. At the same time, this study used a k-means clustering algorithm to cluster 108 students into four categories, and analyzed the differences between four types of students in regard to learning patterns and learning outcomes. The results showed that LMS usage, speed of completing assignments, academic procrastination, peer emotion and social network size were important factors influencing learning outcomes.