Clustering Algorithm to Measure Student Assessment Accuracy: A Double Study

Abstract

Self-assessment is one of the strategies used in active teaching to engage students in the entire learning process, in the form of self-regulated academic learning. This study aims to assess the possibility of including self-evaluation in the student’s final grade, not just as a self-assessment that allows students to predict the grade obtained but also as something to weigh on the final grade. Two different curricular units are used, both from the first year of graduation, one from the international relations course (N = 29) and the other from the computer science and computer engineering courses (N = 50). Students were asked to self-assess at each of the two evaluation moments of each unit, after submitting their work/test and after knowing the correct answers. This study uses statistical analysis as well as a clustering algorithm (K-means) on the data to try to gain deeper knowledge and visual insights into the data and the patterns among them. It was verified that there are no differences between the obtained grade and the thought grade by gender and age variables, but a direct correlation was found between the thought grade averages and the grade level. The difference is less accentuated at the second moment of evaluation—which suggests that an improvement in the self-assessment skill occurs from the first to the second evaluation moment