Bisecting Grid-Based Clustering Approach and Its Validity

A new grid-based clustering approach is presented. By hierarchically bisecting each grid into two volume-equal new grids, this approach can use a new criterion to measure the dissimilarity among all grids and find these candidates of all prototypes. Therefore, the proposed approach can overcome the parameter-sensitive defects in most conventional grid-based clustering approaches, and work well in such dataset with arbitrary-shaped and density-skewed clusters at linear computational complexity. Two experiments are used to verify its clustering effectiveness.