Method for clustering uncertain data

Abstract

A method for clustering uncertain data belongs to the field of data acquiring and processing technology. The invention aims to provide the method for clustering the uncertain data, wherein the method has functions of utilizing an essential latent association of the data, searching a true covariance structure in bottom data recording under an uncertain surface, extracting main characteristics of the data and performing noise reduction clustering. The method comprises the following steps of 1, acquiring the true covariance structure of the bottom data, and 2, according to the covariance structure, performing sharpening noise-reduction processing. The method of the invention can greatly reduce uncertainty of the data from bottom, and furthermore the sharpened noise-reduced data which are obtained through processing can be applied in other fields of fusion, classification, etc. The method has high extendibility.

Description

technical field [0001] The invention belongs to the technical field of data collection and processing. Background technique [0002] In recent years, with the advancement of technology and the deepening of people's understanding of data acquisition and processing technology, uncertain data has been widely valued. In many practical applications, such as economic, military, financial, telecommunications and other fields, data Uncertainty is pervasive and plays a key role. The emergence of uncertain data poses great challenges to traditional cluster analysis ] . The uncertainty of data comes from a variety of situations, the error generated by the data collected by physical instruments, the influence of the surrounding environment on the data in the case of sensor network applications, the bandwidth, transmission delay, energy, etc. The interference of factors and the special purpose of privacy protection may lead to data uncertainty. [0003] The manifestations of data unc...

AI Technical Summary

Problems solved by technology

[0005] To sum up, we can find that most of the existing clustering methods for uncertain data are improved from the traditional clustering algorithms dealing with deterministic data, and they mainly have the following problems: (1) The improved Although the clustering algorithm has improved the clustering quality of dealing with uncertain data, but because it has not substantially reduced the uncertainty of the data, in practical applications, there will still be a phenomenon that the clustering results are seriously affected by errors.

(2) When the improved clustering algorithm deals with high-dimensional uncertain data, the uncertainty aggravates the sparsity of high-dimensional data. Conversely, the sparseness of high-dimensional data will also make the uncertainty of data more difficult. Control processing, so it is not ideal for dealing with high-dimensional uncertain data results

(3) Because the traditional clustering algorithm is improved for the characteristics of uncertain data, the computational complexity and time overhead of many improved algorithms are increased, which greatly reduces the clustering efficiency

Images