Scalable clustering mechanism to analyze the traces and to predict the behavior of learners

Interactive Learning Environment has proven to be an effective solution for overpowering the limits of traditional one-to-many teaching. However, these environments require accurate representation of knowledge about learner to provide suitable guidance. Based on the behavior and performance, the traces of learners are captured and categorized to provide appropriate guidance in the learning environment. The learner networks in E-Learning environment are normally of prodigious, and it is quite difficult to give guidance to each. Thus learners are clustered using traces. When the number of users in the environment increases, scalability comes into an issue. The issue has been solved as organizing the data instances in a way, so that unsupervised clustering algorithm will able to create clusters for ´n´ number of learners. The edges rather than nodes have been taken as data instances. The proposed ECC (Edge centric clustering), treats each edge as one data instance and the connected nodes are the corresponding features. The variance of K-means is applied over the extracted edges, to partition the edges into disjoint sets. This paper contributes to achieve scalable clustering using sparse dimensions. The proposed algorithm has been implemented for the data gathered from VPL integrated with Moodle in Java using Eclipse IDE. The result shows that the proposed approach can efficiently handle huge network while compared to K-means.