JADE 6th edition | Page 103

ARTICLE #6 | 103 DATA MINING FOR LEARNING ANALYTICS including all 66 students who were originally enrolled on the module. The K-means algorithm attempts to find k clusters in a set of observations/samples. Once the algorithm is run and clustering is completed each sample is assigned to the cluster with the nearest centroid (cluster centre). The centroid of a cluster is one that best represents the cluster. The centroid’s attributes are computed by finding the means of the attribute values of the cluster’s members. Results For the programming component of the module there were a total of 2,622 views of related materials (mean=39.7 views per student). The following table shows the breakdown of views for the different material types. Resource Type Number of Files Total Views Avg. Views per Student (n=66) Tutorial Instructions 11 1559 23.6 Lecture Slides 231 825 12.5 Coursework Specification 1 127 1.9 Table 1: Breakdown of views per resource type In order to determine clusters of student behaviour on the module, the following features were then considered: • the student’s degree programme (a code comprised of “P” followed by a set of numbers) • their coursework mark for the programming component of the module • their physical attendance percentage in lectures and tutorials • the number of times they have viewed module related programming materials such as lecture slides, tutorial instructions and the coursework (CW) specification.