Kernel-based possibilistic c-means clustering method of maximum central interval

Abstract

The invention discloses a kernel-based possibilistic c-means clustering method of a maximum central interval. The division problem of the boundary fuzzy data in the high-dimensional space can be solved. An inter-class maximum penalty term is introduced based on the kernel-based possibilistic c-means clustering KPCM algorithm, and the novel kernel-based possibilistic c-means clustering (MKPCM) method of the maximum central interval can be provided. The method provided by the invention is advantageous in that because most of the characteristic vectors are high-dimensional, the good division of the characteristic vectors can be realized by the mapping of the kernel function; and at the same time, the caring between the classes can be considered at the same time, and the optimal division of the boundary fuzzy data set can be realized by adopting the regulation and control of the parameters, and at the same time, the good robustness of the noise points of the KPCM is maintained.

Description

【Technical field】 [0001] The invention relates to the technical field of data mining and pattern recognition, and relates to cluster analysis and image segmentation of data sets. 【Background technique】 [0002] Cluster analysis is an important method in unsupervised pattern recognition, which has been widely used in data mining, image processing, computer vision, bioinformatics and text analysis. The clustering algorithm can classify the data with unknown distribution, find out the structure hidden in the data, and make the data with the same nature belong to the same class as much as possible according to a certain degree of similarity. RuspiniEH. first proposed the concept of fuzzy partition, and introduced fuzzy set theory into cluster analysis. With the introduction of fuzzy theory, people have gradually accepted fuzzy cluster analysis in view of the fuzzy nature of classification. [0003] Fuzzy c-means (FCM) clustering algorithm is one of the most commonly used fuzzy...

AI Technical Summary

Problems solved by technology

The emergence of clustering algorithms based on kernel functions (such as kernel-based fuzzy c-means (KFCM) clustering algorithm and kernel-based likelihood c-means (KPCM) clustering algorithm) has overcome FCM, PCM and related problems to a certain extent. The improved algorithm is not suitable for the defects of various data structures such as non-hyperspheres, but they still have the defects of the original algorithm, that is, the distance relationship between the class centers is ignored, when the ambiguity at the boundary is high or there are noise points at the boundary When performing cluster analysis with the data set of outliers, it is easy to produce the phenomenon that the positions of the cluster centers overlap or shift, so that the data at the boundary may be misclassified.

Images