Analyzing Student Programming Paths using Clustering and Process Mining

Abstract

Learning programming is becoming more and more common across all curricula, as seen by the growing number of tools and platforms built to assist it. This paper describes the results of an empirical study that aimed to better understand students’ programming habits. The analysis is based on unsupervised classification algorithms, including features from previous educational data mining research. The k-means method was used to identify the behaviors of six students profiles. The main and interaction impacts of those behaviors on their final course scores are tested using analysis of covariance.