Visualization of Programming Skill Structure by Log-Data Analysis with Decision Tree

To evaluate the ability of a programmer in universities or staff agencies, evaluators conduct examination in which applicants solve problems, or they refer to the applicants’ GitHub and review their code. However, the collected source code is evaluated by a human under present conditions. This leads to two problems: many source codes cannot be evaluated simultaneously, and it is difficult to maintain consistency in the evaluation criteria among evaluators. We propose methods to estimate the current skill of the programmer, which can be used by the evaluators to understand an applicant’s skill by analyzing the collected source code automatically. In particular, this study visualizes the feature of the students’ source code in term of quarters using a decision tree to estimate the programming skill.