Self-adaptive rapid K-means clustering method fusing feature learning

Abstract

The invention discloses a self-adaptive rapid K-means clustering method fusing feature learning. The method comprises the steps: firstly, preprocessing data, eliminating the problems of attribute deficiency and data repetition, and conducting normalization processing on all data attributes; calculating a data total divergence matrix, and introducing sparse characteristics to construct a feature selection matrix; performing K-means clustering method on feature subspace, introducing an adaptive factor to dynamically adjust the weight of each data sample in the updating process of a clustering center; and updating the feature selection matrix according to distinguishable information between the clusters, and further screening out an optimal feature subset. The method makes the traditional K-means clustering method efficiently use distinguishable information between the clusters and in the clusters and correlation information between the features to improve the clustering accuracy. In addition, adaptive factors are fused in the clustering process, the clustering center is updated according to the distribution characteristics of different types of data, high practicability and expandability are achieved, and effective support can be provided for machine learning, computer vision and other related applications.

Description

technical field [0001] The invention belongs to the technical field of machine learning, in particular to an adaptive fast K-means clustering method for fusion feature learning. Background technique [0002] Clustering method is a technology widely used in the field of machine learning, among which the K-means method is the most widely used, and it has achieved good results in various fields such as data mining, medical treatment, and education. However, with the rapid development of multimedia technology and Internet technology, the explosive growth of high-dimensional data has brought great challenges to the traditional K-means method. Due to redundant features and noise features in high-dimensional data, directly applying K-means clustering to such data not only consumes a lot of computing resources, but also affects its clustering accuracy. The latest research shows that if the data features are dimensionally reduced in advance, the clustering efficiency of K-means will...

AI Technical Summary

Problems solved by technology

Although such methods can effectively improve the clustering accuracy of K-means, they all need to use the eigendecomposition operation to solve the optimal feature subspace, and its computational complexity will increase with the quadratic level of the feature dimension of the data to be processed, and the obtained The feature subspace is quite different from the original feature space, and it is difficult to apply to high-dimensional data processing in real application scenarios

Images