Clustering Analysis of Student Learning Outcomes Based on Education Data

This Research to Practice Full Paper extracts knowledge from the education data through clustering student outcomes for application towards refining course design. Changes in the field of technology and other aspects of the work environment calls for continual advancement in the education sector. Analysis of the outcomes of students from the Senior Exit Survey gauges the competence of graduates in their field of study. This survey is routinely conducted on students close to finishing an undergraduate degree and contains information about the various experiences while learning at Cal Poly Pomona. Understanding of these outcomes allows the faculty and the administration to improve the courses in an effective manner. The basis of this study is to cluster student learning outcomes and distinguish those with superior similarity. These outcomes can be fused for a targeted approach towards designing optimized courses.The clusters were developed using one of the most frequently used unsupervised learning techniques, or, the hierarchical clustering algorithm through R statistical analysis software. This algorithm had been proven to be reliable yet easy to interpret in past literature. For the purposes of this study, the hierarchical clustering model is defined by the dissimilarity measure between each pair of observation using the Euclidean distance along with the use of both complete and average linkages. The results from the two linkages were displayed using dendrograms. For additional visualization and verification of the clusters, a heat map was also constructed to illustrate the results using the complete linkage. A comparison of the results from these two linkages demonstrates exceptional similarities amongst the clusters; all but one outcome did not fall within the same clusters. In conclusion, the results show there exist three major clusters and three pairs of closely related outcomes to form out of the Senior Exit Survey data.