HYPER-SPECTRAL DATA CLUSTERING METHOD BASED UPON THE SENSITIVE SUBSPACE

We put forward a quick clustering method for large numbers of data, with high dimensions, which is based on sensitive subspace consisting of the data set's sensitive dimensions. We first estimate the probability density of each dimension by the parzen window algorithm, enhance its optional ability through extracting zeroes and smoothness processing, then through the recognition of the number of the rallying points and the gain of the sensitive dimensions in order to compose the sensitive subspace, and lastly, we perform the Rival Penalized Competitive Learning (RPCL) clustering in the subspace. Moreover, we detected the red tide of hyper-spectral data using this method. Furthermore, the overall improvement in terms of the computational speed is about nine times faster.