Nonlinear manifold learning dimension reduction method based on adaptive density clustering

Abstract

The invention discloses a nonlinear manifold learning dimension reduction method based on adaptive density clustering. The method comprises the following steps: 1) after the adaptive density clustering algorithm is used for clustering, ICA (Independent Component Analysis) is used for carrying out dimension reduction on each cluster to form a linear model plane; 2) the minimum spanning tree (MST) between local linear models is built; 3) manifold global MSTs are transversed; and 4) through operating the ICA on a global hyperplane, low dimensional implantation is found out. According to the nonlinear manifold learning dimension reduction method based on adaptive density clustering, the parallel mapping of the plane is used for overcoming the distortion generated by dimension reduction of the original data set, the accuracy is high, and the credibility is good.

Description

technical field

[0001] The present invention belongs to a nonlinear data dimensionality reduction method. Aiming at the difficulties in directly carrying out data mining and analysis on high-dimensional data in current big data applications, a nonlinear popular method based on adaptive density clustering is proposed by using a parallel mapping method. Learning the dimensionality reduction method and using the parallel mapping of the plane can overcome the distortion of the original data set due to dimensionality reduction. Background technique

[0002] With the development of science and technology and the era of big data, data information is rapidly changing to high-dimensional. The data information generated by people in the course of behavior is no longer a simple small amount of data, but high-dimensional data containing a large amount of information, but the huge The amount of data and the high-dimensional eigenvalues ​​of each data sample bring difficulties to data pro...

Images