Dynamic Clustering and Indexing of Multi-Dimensional Datasets

Abstract – Nowadays large volumes of data with high dimensionality are being generated in many fields. Most existing indexing techniques degrade rapidly when dimensionality goes higher. ClusterTree is a new indexing approach representing clusters generated by any existing clustering approach. It is a hierarchy of clusters and subclusters which incorporates the cluster representation into the index structure to achieve effective and efficient retrieval. The authors propose a new ClusterTree indexing structure which has new features from the time perspective. Each new data item is added to the ClusterTree with the time information which can be used later in the data update process for the acquisition of the new cluster structure. This approach guarantees that the ClusterTree is always in the most updated status which can further promote the efficiency and effectiveness of data insertion, query and update. This approach is highly adaptive to any kind of clusters.