Mixed attribute data set clustering method for automatically determining clustering center

Abstract

The invention designs a mixed attribute data set clustering method for automatically determining a clustering center. Aiming at the problem of low algorithm accuracy and stability caused by manually specifying an initial clustering center and a clustering number in a K-Prototypes clustering algorithm, the method comprises four steps of initializing, pre-selecting clustering center points, determining the clustering center points and iterating a clustering division process. In the invention, based on density distribution of data objects, automatic identification of the number of class clustersis realized, an initial clustering center is selected, a local optimal problem caused by initial point selection is optimized, in addition, by distinguishing different influence weights of each attribute on a clustering result and improving a dissimilarity calculation formula, the clustering accuracy is improved, and a better clustering effect is achieved.

Description

technical field [0001] The invention belongs to the field of data mining, and in particular relates to a hybrid data set clustering method including numerical attributes and classification attributes. Background technique [0002] As an important part of the data mining field, clustering can discover the underlying structure and organizational rules of unlabeled data. In practical applications, data information not only includes quantitative numerical attributes such as age and height, but also category attributes such as gender and occupation. Mixed datasets containing both numeric and categorical attributes are ubiquitous. At present, the K-Prototypes clustering method is mostly used to solve the clustering problem of mixed data. This method has the characteristics of simplicity and high efficiency, and is widely used, but there are some shortcomings. The random selection of the initial clustering center in the K-Prototypes clustering algorithm may cause different conver...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problem of low accuracy and stability of the algorithm due to the need to manually specify the initial cluster center and the number of clusters in the K-Prototypes clustering algorithm, the present invention proposes a clustering method for mixed attribute data sets that automatically determines the cluster centers, According to the density of data objects and the distance between data objects, this method realizes the automatic identification of the number of clusters, selects the initial cluster center, optimizes the local optimal problem caused by the initial point selection, and ensures the accuracy of the clustering results

Images