Learners need design abilities for programming. It is necessary for them to have prospects indicating what programming knowledge they organize into a code. Since programming knowledge to be used in each assignment has been decided in advance, careful consideration is necessary to make learners to achieve design abilities through assignments. If the knowledge is specified in assignments, programming exercise courses cannot train learners to achieve design abilities. Meanwhile, if the assignments do not specify it, learners without programming prospects may be at loss what they should begin with. Supervisors should discriminate those learners from others in programming exercise courses. This paper proposes a method to determine whether learners work on assignments along with programming prospects. The method acquires reference histories to programming knowledge learners consult to get programming prospects. A classifier trained with a machine learning algorithm discriminates reference histories with programming prospects from others. We verify the usefulness of the method, comparing the result using the method with what is manually derived from source codes. From this result, reference history for a specific time before this time can find programming prospect of learners. Supervisor cannot only grasp programming prospects of learners, but also provide suitable advices for individual learners, monitoring the transition of programming prospect.

Full text: IJEIR-1394_final.pdf