A New Elliptical Grid Clustering Method

AbstractA new base on grid clustering method is presented in this paper. This new method first does unsupervised learning on the high dimensions data. This paper proposed a grid-based approach to clustering. It maps the data onto a multi-dimensional space and applies a linear transformation to the feature space instead of to the objects themselves and then approach a grid-clustering method. Unlike the conventional methods, it uses a multidimensional hyper-eclipse grid cell. Some case studies and ideas how to use the algorithms are described. The experimental results show that EGC can discover abnormity shapes of clusters