Unification of Subspace Clustering and Outliers Detection On High Dimensional Data

With the unanticipated requisites springing up in the data mining sector, it has become essential to group and classify patterns optimally in different objects based on their attributes, and detect the abnormalities in the object dataset. The grouping of similar objects can be best done with clustering based on the different dimensional attributes. When clustering high dimensional objects, the accuracy and efficiency of traditional clustering algorithms have been very poor, because objects may belong to different clusters in different subspaces comprised of different combinations of dimensions. By utilizing the subspace clustering as a method to initialize the centroids, and combine with fuzzy logic, this paper offers a fuzzy subtractive subspace clustering algorithm for automatically determining the optimal number of clusters. By our new Fuzzy Outlier detection and ranking approach, we detect and rank the outliers in heterogeneous high dimensional data. The experiment results show that the proposed clustering algorithm can give better cluster validation performance than the existing techniques.