Abstract:

Evaluating a student’s work is the first, all-important step in giving feedback. However, evaluating the quality of students’ work is usually a time-consuming process. This is further complicated by the fact that it is necessary to consider many factors that affect the assessment, such as the time spent on completing the task, the number of attempts to complete the task, the level of complexity of the completed task, etc.

Automating the assessment would undoubtedly accelerates this process. Machine learning methods are increasingly being applied to analyse data used to evaluate students. In the paper, we analyse and show how a decision tree method can be used to automate the process of assessing students. We have created sets of assessment questions for two learning courses: Programming Fundamentals and Object-Oriented Programming, within one curriculum of the Virumaa College at Tallinn University of Technology.

The results of student performing the assessments were input data for various machine learning algorithms that allow for forming decision trees. Bloom’s well-known taxonomy was used to compile these sets of assessment questions. An analysis of decision trees was performed to check how this method can be used to optimise evaluating the results of students’ work. The obtained results and conclusions allow us to analyse how the evaluation of the student’s work results can be presented in a conceptual model of personalised feedback for students’ work, thereby creating a software reference architecture to automate the evaluation of students’ work.

Keywords:

Grade, Bloom’s taxonomy, machine learning, decision tree.