An improved grid-based k-means clustering algorithm

• DOI: 10.14257/astl.2014.73.01
• Publication Date: 2016-03-04T00:00:52Z
• Authors: Li Ma; Lei Gu; Bo Li; Linlin Zhou; Jin Wang
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

The traditional K-means clustering algorithm is difficult to initialize the number of clusters K, and initial cluster centers are selected randomly, this makes results very unstable. Meanwhile, algorithms susceptible noise points. To solve problems, improved. improved method divided into same grid in space, according size data point property value assigns it corresponding grid. And count points each Selecting M (M>K) grids, comprising maximum points, calculate central point. These as input data, then determine k based on results. In find K farthest from other those center algorithm. At time, must be included K. If less than threshold, these will considered removed. Theoretical analysis experimental show that compared has high quality results, iteration good stability.