Weighting contraction method based on K near neighbor method

Abstract

The invention relates to a weighting contraction method based on a K near neighbor method, which comprises the following steps that: samples are divided to intensively belong to training samples and test samples in each classification; the sample point contraction for preset times is respectively carried out on the training samples in each classification for obtaining the training set data corresponding to each classification, and the weight of each sample point after the contraction is calculated; and the test samples are classified according to the weights of K sample points in the training set data with the nearest distance away from the test samples. The invention can ensure the accuracy of the classification at the same time of considering the computer calculation speed improvement and calculation complexity reduction.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a weighted shrinkage method based on the K-nearest neighbor method. Background technique [0002] In the past fifty years, the theory and technology of pattern recognition and machine learning have been developed rapidly. As a typical classification method of pattern recognition, the nearest neighbor method was first proposed in 1968. Although the classification result of this method is not the best, its computational complexity is small and easy to implement, so it has been widely used. The k-nearest neighbor method is a direct expansion method of the nearest neighbor method. In order to reduce the amount of calculation and storage of the k-nearest neighbor method, a lot of research has been carried out, and methods such as the clipping nearest neighbor method and the compression nearest neighbor method have been produced. Although these methods reduce the computation...

AI Technical Summary

Problems solved by technology

Although these methods reduce the complexity of operations to a certain extent, they also have a certain negative impact on the classification effect.

In addition, the sample points in each classifier in these methods are at the same level, so the role of more important samples in the classification process of these methods cannot be reflected

Images