K-means initial clustering center optimization method on basis of neighborhood information and mean difference degree

Abstract

The invention discloses a K-means initial clustering center optimization method on the basis of neighborhood information and the mean difference degree. The method includes the steps that 1, a sampleset X={X1, X2, ..., Xi, ..., Xn} with n objects is input, the clustering number K is determined, and the current determined initial clustering center number k=0 is initialized; 2, a distance matrix Dis formed; 3, the neighborhood radius value delta is determined; 4, the number Ni of samples in the delta neighborhood of each sample point Xi is calculated, and a matrix N is formed; 5, the sample point Xi corresponding to the maximum sample number Ni in the delta neighborhood in N is regarded as the first clustering center C1, k=k+1, and the corresponding Ni in N is set as 0; 6, the sample pointXj corresponding to the maximum sample number Nj in the delta neighborhood in N is searched for, the distances between Xj and the clustering centers {C1, C2, ..., Ck} are calculated, and the corresponding Nj in N is set as 0; 7, if the distances between Xj and the clustering centers are not less than the mean difference degree M, k=k+1, C(k+1)=Xj, or Step 6 is returned to; 8, if the current clustering center number k is equal to the clustering sort number K, K initial clustering centers are output, or Step 6 is returned to; 9, the whole sample set is clustered by means of the K-means clustering algorithm, and a clustering result is output.

Description

technical field [0001] The invention relates to the technical field of data mining, in particular to a K-means initial clustering center optimization method based on neighborhood information and average difference degree. Background technique [0002] Clustering algorithm is an unsupervised classification algorithm, which refers to dividing a group of objects without category identification into several categories according to a certain similarity, so that the distance between objects between categories is as large as possible, and the distance between objects within a category is as small as possible. One of the basic methods of system modeling and data mining has been widely used in various fields, such as text classification, image recognition and other fields. [0003] The K-means clustering algorithm (K-means) is a partition-based dynamic clustering algorithm. Because of its simplicity, it has become one of the most popular clustering methods. The disadvantages of the ...

AI Technical Summary

Problems solved by technology

The disadvantages of the traditional K-means clustering algorithm are as follows: First, the clustering results of the algorithm are easily affected by the initial clustering center, and when the initial clustering center is unreasonably selected, there will be consistent clustering and failure to converge

Second, the adverse effects of outliers on the clustering results cannot be overcome, resulting in unstable clustering results and low accuracy

Images