Mixed feature data clustering method and system based on tree base learner

Abstract

The invention belongs to the technical field of mixed feature data set clustering, and discloses a mixed feature data clustering method and system based on a tree base learner, and the method comprises the steps: S1, carrying out the random sub-sampling of a sample set, and generating N different sub-sample sets; s2, performing tree-based learning device training on each sub-sample set, and obtaining N trees after training and the number K of clusters; s3, counting a similarity matrix between any two samples based on the N trained trees, and normalizing all the similarity matrixes to obtain a plurality of normalized similarity matrixes; and S4, taking the number K of the clusters and the plurality of normalized similarity matrixes as input of a spectral clustering model to obtain a final clustering result of the sample set. The data clustering method under the conditions of high latitude and mixed features is designed, and the problem that clustering is difficult due to the fact that a similarity concept cannot be clearly defined under the conditions that the data set dimension is too high and continuous features and discrete features are mixed can be solved.

Description

technical field [0001] The invention belongs to the technical field of clustering of mixed feature data sets, and in particular relates to a method and system for clustering mixed feature data based on a tree-based learner. Background technique [0002] For the vehicle data set of the energy supply station, the high dimensionality of data features and the mixture of continuous and discrete features pose challenges to traditional clustering algorithms, especially clustering algorithms based on Euclidean distance. In the case of the "curse of dimensionality", all samples will be approximately equidistant and adjacent, so that the nearest neighbor problem may become meaningless. Second, many algorithms that rely on traditional distance measures are sensitive to attributes of different units. Although data transformation can be used to alleviate this problem, it may change the distribution of data and affect the clustering results. Moreover, in the case of large data, an excess...

AI Technical Summary

Problems solved by technology

This patent uses Euclidean distance for clustering, which cannot accurately cluster high-latitude mixed data

Images