A gird-based fuzzy cluster approach

A new fuzzy clustering approach is presented based on two steps: data reduction and core data aggregation in a reduced subset of the original dataset. The data reduction largely reduces a number of data points in a dataset and simultaneously improves clustering quality based on a grid-based initialization for data space, where each grid is continuously bisected into two volume-equal smaller grids, so that a group of core points is found. By clustering these core points, all cluster prototypes are determined. The new approach can work faster and more effective in a dataset when it is compared with most of the existing fuzzy clustering approaches, effectively approximating the number of clusters. Two experiments were used to verify the usefulness of the new approach.