Uncertain Data Clustering Based on Probability Distribution in Obstacle Space.

In the case of current technology, most of the measurements are focused on geometric distance, and the distribution of data is not considered. In order to compensate for this shortcoming of geometric distance measurement, this paper uses the KL distance as the similarity measurement standard for uncertain data, and the DOUD_C algorithm and COUD_C algorithm are proposed respectively in the discrete domain and continuous domain. In order to solve the problem of efficient clustering of the high dimensional data, this paper considers the data structure of grid, and BROUD_C algorithm is proposed. According to the adjacency characteristic of the grid, the cluster process is extended continuously, the algorithm can find clusters of arbitrary shapes, and we can filter a large number of isolated points, it solves the uncertain data clustering problem effectively in the obstacle space. The experimental results show that compared to the OBS_UK_means with VPA and SDA pruning algorithm and FOPTICS algorithm, the clustering performance of BROUD_C algorithm is more significant and CPU has less execution time in the obstacle space. [ABSTRACT FROM AUTHOR]