Grouping and Labeling in Digital Learning Portfolios

This paper describes an application of clustering and labeling methods in actual learning records of e-portfolios implemented through personal blogs. In the clustering process, we have used k-means algorithm with Euclidian distance and random initialization of the centroids, using the elbow method to determine the best value of k; we used centroid labels for labeling. Analysis of 4,241 instances of the model (posts) have shown that absence of formal descriptors was overcome through representation of the titles given to posts, suggesting its use for automatic structuring of contents of this nature.