Estimating Academic results from Trainees’ Activities in Programming Exercises Using Four Types of Machine Learning

Abstract

Predicting trainees’ final academic results in the early stage of programming class is a significant mission in the field of learning analytics. Performing exercises in programming class is hard and it takes a lot of time for trainees. For this reason, careful support with trainees are offered in many classes through classroom assistants (CAs). Even with CAs’ assistances, managing a programming class is difficult. Because each trainee’s coding activity is different from another’s, even when each of them is solving the same exercise. Classroom assistants with little teaching experience have difficulty for understanding the unique features of trainee’s coding activity. We have employed data mining to analyze trainees’ coding activities to distinguish those various features. The objective of this research is to present such behavioral features of trainees to CAs to enrich their assistance for the trainees. In order to investigate the timing of guidance, we conjectured the academic results from the chronicle record of coding activities.