Analysing graduation project rubrics using machine learning techniques

Abstract

When grading a student's performance, determining the assessment factors is a substantial step in course evaluation. The aim of this paper is to improve the quality of the assessment criteria for our Computer Engineering Department's graduation reports. We employ machine learning methods to identify the most important evaluation rubrics that affect the overall grade given to graduation projects. First, we eliminate the redundant factors by computing the correlations between them. Second, we apply K-Means & Hierarchical Clustering methods and third, we analyze the proportion of variance values to find the sufficient amount of eigen values to explain the data. Our results show that Overall Performance is the most important, whereas References is the least important evaluation rubric affecting the graduation project grades. The techniques we use can be used to analyze the graduation rubric grading practices and also to come up with an equivalent rubric with smaller set of questions.