Proposal of a System Enabling Adaptive Support Based on Transition of Each Learner's Source Codes

The difficulties that students face in programming education can be broadly classified into two types: difficulty in conceiving a solution and difficulty during the actual coding process. In this study, we developed a machine learning model that enables adaptive instruction by estimating a solution that will be easy for the learner to understand according to the learner's past work on previous tasks. Our evaluation showed that the developed approach could estimate easily understandable code instances for learners with high accuracy (97.5 % AUC).