Log Analysis Using Bayesian Networks in a Card Operation Based Programming Learning Support System

The Card Operation Programming Study Support System (COPS) was developed to help beginning students understand the structure of programs.COPS is a card-based programming method in which learners solve problems and create programs by rearranging cards according to problem statements.By sorting cards according to the problem statement, learners can visually understand the structure of the program needed to solve the problem.It is expected to have the same learning effect as the conventional coding format, and its educational effect has been suggested from the perspective of Learning Analytics (LA), which utilizes learning history data.In particular, a method for estimating learners ' thinking patterns by analyzing COPS usage logs using a Bayesian network (BN) has been proposed and its usefulness has been partially confirmed.However, several issues remain in the conventional LA approach, and factors such as the constraints of the probrems faced by the learner have not been adequately taken into account.Therefore, in this study, we propose a method to analyze COPS learning history data with BN to visualize how learners tackle problems in more detail.Specifically, we extract the correct and incorrect patterns of learners ' card operations from the logs and use BN to learn their structure in order to estimate how learners tackle and understand the problems.In our experiments, we analyzed the data obtained for basic problems in the C language.The results showed that certain card manipulation patterns were associated with incorrect answers, confirming that analysis using BN is effective in supporting learners.