Predicting programming performance by using process behavior in a block-based programming environment

In this study, a predictive model is constructed to analyze learners' performance in programming tasks using data of programming behavioral events and behavioral sequences. First, this study identifies behavioral events from log data and applies lag sequence analysis to extract behavioral sequences that reflect learners' programming strategies. Then, learners' behavioral events and behavioral sequences are selected using linear regression. Finally, a predictive model is built using majority vote. The study uses data from 3151 programming tasks to compare the accuracy of predictive models with and without behavioral sequences. The results show that the accuracy of the predictive model with behavioral sequences is 75.75%, which is 2.06% higher than that of the model without behavioral sequences. This study finds that editing, debugging, and self-regulating behavior sequences can reflect the strategies learners employ during programming. These findings underscore that the model based on behavioral events and behavioral sequences in this study can successfully predict learners' programming performance and thereby has the potential to support programming instruction.