Abstract: College course study is an important part of improving academic level. Predicting course performance of college students and warning students who have the risk of course failure have become an important topic in the study of big data in education. Electronic check-in system can collect information such as student attendance time and seat selection in classroom. This information reflects the students?? interests in studying, achievement motivation and personality characteristics, and has a strong correlation with course performance. This paper studies the prediction methods of course performance based on electronic check-in data, and builds a performance prediction model combined with the psychological test data of the students, which includes seat selection and attendance time attribute construction, prediction method design, and prediction correction modules. This paper proposes the partition method of classroom seating based on score distribution, and the correction method of partnership effect, which improve the accuracy of performance prediction. Using the real datasets, the proposed prediction model is fully verified by experiments, and the average prediction error of hundred-mark is less than 10.

Key words: electronic check-in, feature selection, personality characteristics, performance prediction

摘要： 大学生的课程学习是提高学业水平的重要组成部分。预测大学生的课程成绩，预警有课程学习失败风险的学生，成为教育大数据研究中的一个重要课题。电子签到系统收集了学生在课堂上签到时间和座位选择等信息。这些信息反映了学生的学习兴趣、成就动机与性格特征等因素，与课程成绩具有较强的相关性。研究了基于课堂电子签到数据的课程成绩预测方法，结合学生的心理测试数据构建了成绩预测模型，模型中包括座位选择与签到时间等属性构造、预测方法设计、预测结果修正等模块。提出了基于成绩分布的教室座位分区划分方法和同伴影响的预测结果修正方法，提高了成绩预测的精度。利用真实数据集对所提出的预测模型进行了充分的实验验证，百分制成绩平均预测误差在10分以内。

关键词: 电子签到, 属性选择, 个性特征, 成绩预测