Student Course Score Prediction Based on Large-scale Multi-Source Data

With the rapid development of education informatization, many scholars researched in the field of education data mining (EDM), among which the most popular research is student performance prediction. Early research of student performance prediction mainly relied on students' performance data. With the advent of the big data era, available data for student performance prediction has been greatly expanded. However, the main problems existing in the present study of performance prediction are the simple data source and small data scale. To solve these problems, this paper studies on performance prediction based on large-scale multi-source data. This paper creates a large dataset - GLMS (Grade-Point Average with Learning Management System Dataset) by using the data of students' scores and learning behaviors in a secondary college of a university. This dataset contains the data of 1153 students in the same grade in three years, which covers the data of 15 courses that should be taken. Besides, this paper includes preprocessing of input data, feature extraction, comparison of prediction algorithms and evaluation of prediction results.