Data Streams Clustering Algorithm Based on Grid and Particle Swarm Optimization

The offline components of CluStream clustering algorithm based on distance, and it is difficult to find non-spherical character of the cluster. This paper proposes a data streams clustering algorithm based on grid and particle swarm optimization, the algorithm based on two-tier structure of CluStream clustering algorithm. The grid feature vector to represent a snapshot, the grid density, and grid merging technologies is applied in this algorithm. So we can found any non-spherical clusters, In this algorithm. Non-dense grid is periodic and dynamic way to remove, which is good for reducing the space complexity. Using the PSO optimized clustering results in the offline components, in order to get a more precise clustering efficiency. Experiments show that this algorithm is more efficient than the CluStream algorithms, it has a good number of dimensions scalability, and it can find non-spherical nature of the clustering results.