Rich Kid, Poor Kid: Students Clustering using K-Prototype Algorithm

This article attempts to utilize a clustering-based analysis on student data in an attempt to gain insights and help improve student learning outcomes and achievements. In doing so, the k-Prototype clustering algorithm was used on Open University’s Learning Analytics Dataset which contained information on the student’s socio-economic background, interactivity levels with their virtual learning environment, and academic performance. Through the use of the elbow method, it was discovered that the most optimum number of clusters to group students were two clusters. Both clusters’ sample were analyzed, and was then labeled and compared. The first cluster was labeled as the “privileged group”, as it represented students that lived in areas that were less deprived and performed well academically. On the other hand, the second cluster was labeled the “less-privileged group”, as students in this cluster lived in areas with higher numbers of indices of multiple deprivation (IMD) and lesser performance. This study indicates the suitability of the k-Prototype clustering algorithm to cluster students, and implies for future research to use more diverse datasets to discover more insights.