Compact clusters on topic-based data streams

Abstract

The regular clustering algorithms follow a batch mode of clustering. The existing clustering algorithms that use the incremental approach are lagging with drawbacks like inefficient memory utilization, data loss. The proposed system resolves all the issues by incremental clustering algorithm to cluster data incrementally. The clustering algorithm uses heuristic measures. Every incoming data is compared with the existing clusters. If it matches with any of the existing clusters, it finds a place to reside. Otherwise, it creates a new cluster. There are a few prerequisites. The raw data (unstructured) has to be brought into a structured format to initiate the clustering process. Each cluster is represented in the form of a vector called a data cluster vector. Finally, the proposed method is proved to be better than the existing method, and misclassification is completely removed.