An Interpretable Online Learner's Performance Prediction Model Based on Learning Analytics

Abstract

Most of student performance prediction model only focused on the accuracy of prediction results, but achieving an interpretable prediction model may be as important as obtaining high accuracy in learning prediction research. This paper proposed a student performance prediction model based on online learning behavior analytics with 19 behavior indicators. This model consists of four steps: data collection and processing, correlation analysis, data analytics, student performance prediction algorithm, prediction and intervention. Moreover, a case have been taken to predict student performance according to the model with rule-based genetic programming algorithm. The experiment results show that the rule-based genetic programming algorithm has a stronger interpretation in ensuring competitive prediction accuracy. The model achieves a good prediction effect.