Search

Clustering high-dimensional data stream is a difficult and important problem. Grid-based algorithms are easily influenced by the size and borders of the grid. To overcome the weakness, we propose a new Irregualr Grid-based Clustering algorithm for high-dimensional data streams, called IGDCL. This method incorporates an irregular grid structure and subspace clustering algorithm. In this paper, an irregular grid structure is generated by means of splitting each dimension into different grid cells. With new data arriving, the irregular grid structure is dynamically adjusted. We assign a fading density function for each data point to embody the evolution of data streams. The final clusters are obtained in subspaces which are formed by dimensions associated with corresponding clusters. Experimental results demonstrate that IGDCL has higher clustering quality than CluStream.

Clustering high-dimensional data stream is a difficult and important issue. In this paper, we propose MStream, a new clustering algorithm based on matrix over high dimensional data stream. MStream algorithm incorporates a synopsis structure, called GC (Grid Cell Structure), and grid matrix technique. The algorithm adopts the two-phased framework. In the online component, the GC is employed to monitor one-dimensional statistics data distribution of each dimension independently. Sparse GCs which need to be deleted are checked by predefined threshold. In the offline component, it is possible to tracing multi-dimensional clusters by dense GCs which are maintained in the online component. Grid matrix technique is introduced to generate the final multi-dimensional clusters in the whole data space. Experimental results show that our algorithm has the flexible scalability and higher clustering quality.