Clustering Methods For Spatial Datamining

Abstra t We investigate the use of biased sampling a ording to the density of the dataset, to speed up the operation of general data mining tasks, su h as lustering and outlier dete tion in large multidimensional datasets. In density-biased sampling, the probability that a given point will be in luded in the sample depends on the lo al density of the dataset. We propose a general te hnique for density-biased sampling that an fa tor in user requirements to sample for properties of interest, and an be tuned for spe i data mining tasks. This allows great exibility, and improved a ura y of the results over simple random sampling. We des ribe our approa h in detail, we analyti ally evaluate it, and show how it an be optimized for approximate lustering and outlier dete tion. Finally we present a thorough experimental evaluation of the proposed method, applying density-biased sampling on real and syntheti data sets, and employing lustering and outlier dete tion algorithms, thus highlighting the utility of our approa h.