Method for realizing clustering and clustering boundary defining of real-time data streams with noise points

Abstract

The invention particularly relates to a method for realizing clustering and clustering boundary defining of real-time data streams with noise points. The scheme includes that density of data points capable of mapping to grids g is updated; the grids g with densities higher than or equal to density threshold (t) are marked as high-density grids gh; adjacent high-density grids gh or low-density grids gl' which have densities higher than similar threshold are marked as clustering grids ggrid; all the clustering grids ggrid are clustered; grids gl' which are not clustered and are adjacent to the clustering grids ggrid or the clustering grids ggrid which are positioned at the edge of a grid space are marked as clustering boundary grids gboundary, and all the clustering boundary grids gboundary form clustering boundaries; all the clustering and clustering boundaries are outputted; next clustering request time tnext is calculated; and the above steps are repeated for a data stream at the moment of the time tnext, until the data stream D is operated. The method has the advantages of high clustering efficiency and fine boundary defining effect.

Description

technical field [0001] The invention belongs to the technical field of data processing of data flow. Specifically, it relates to a method for clustering and clustering boundary definition for real-time data streams containing noise points. Background technique [0002] The method of clustering and defining cluster boundaries for real-time data streams containing noise points can improve the accuracy of clustering and data classification, and can quickly discover clusters and cluster boundaries in real-time data streams, and can also be used for industrial production. Monitoring of real-time production process data in the system is helpful for production equipment and product quality monitoring. At present, the definition and detection technology of data clustering boundary points has the following defects: (1) the existing algorithm only extracts the boundary points that define clusters for static data sets, and there is no method for defining cluster boundaries for real-ti...

AI Technical Summary

Problems solved by technology

At present, the definition and detection technology of data clustering boundary points has the following defects: (1) the existing algorithm only extracts the boundary points that define clusters for static data sets, and there is no method for defining cluster boundaries for real-time data streams; (2) Separate clustering and boundary detection and process them separately; (3) The complexity of the algorithm is high in large data processing

The BORDER algorithm can detect the boundary points of clusters in the data set without noise points, but its disadvantages are: (1) It cannot correctly identify the boundary points in the data set containing noise, because the reverse k-nearest neighbors of noise points The number is less than the number of reverse k-nearest neighbors of clustering boundary points; (2) The algorithm needs to find the k nearest neighbors of each object, and then calculate the number of reverse k-nearest neighbors of each object, The execution efficiency of the algorithm is not high; (3) the user needs to have prior knowledge, and the number n of boundary points of the data set is given

Images