Improving Students’ Achievement Prediction in Blended Learning Environments with Integrated Machine Learning Methods

Blended learning has been adopted in universities for over a decade. With the advantage of combining online and offline learning contexts, blended learning enables a teacher to design separate instructional models in one curriculum. However, student data from two learning contexts also bring challenges for teachers. Especially, it is difficult to apply machine learning methods to improve students’ learning efficiency. Moreover, there are more obstacles to interpreting results from machine learning methods and transforming them into student-supporting strategies. This chapter discusses a case study using machine learning methods to analyze learning data from both online and offline contexts. The study initially measures students’ learning styles, collects students’ online learning behavior data and offline classroom voice data, and uses an integrated machine learning method to recognize students’ data patterns. Periodic achievement tests are arranged in teaching practice to select the students whose academic performance increment is larger than the average increment of the whole class, and the implication of data patterns of selected students is used to modification of teaching strategies. The results indicate that after at least three semesters of iteration, about 79% of students in the class follow the recommended learning pattern. Compared to collaborative learning, this method is more suitable for discussion-based learning.