Reference point K-nearest neighbor classification method based on MPI (Message Passing Interface) parallelization

Abstract

The invention relates to a reference point K-nearest neighbor classification method based on MPI (Message Passing Interface) parallelization, and belongs to the field of data classification. The method comprises the following steps that: S1: on the basis of a reference point K-nearest neighbor algorithm, utilizing a distance between a sample point and a plurality of reference points to measure a position difference between the sample point and the reference points, defining a position difference factor, calculating a similarity between a training sample and the reference point, and generatingan ordered similarity sequence; S2: according to the similarity between the training sample and the reference point, searching a nearest neighbor sample in the training set from the ordered sequence;and S3: calculating a definite similarity with the testing sample from the searched nearest neighbor sample so as to find k neighbor samples and judge categories. By use of the method, in virtue of reference points, the searching of the k neighbor is quickened, an MPI technology is used for realizing parallelization, and therefore, the classification speed of large-scale high-dimension data is quickened.

Description

technical field [0001] The invention belongs to the field of data classification and relates to a reference point k nearest neighbor classification method based on MPI parallelization. Background technique [0002] Classification is an important technology in the field of data mining. Its purpose is to construct a classification model (also known as classification function, classifier) ​​according to the characteristics of the data set, which can map samples of unknown categories to certain categories in a given category. one or several. The k-nearest neighbor algorithm was originally proposed by Cover and Hart in 1968. It is a non-parametric classification technique that has the advantages of robustness, clear concept, and easy implementation, and can achieve high classification accuracy for unknown and non-normal distributions. Rate. [0003] The traditional k-nearest neighbor algorithm has a high time complexity. At present, many scholars have proposed many improved alg...

AI Technical Summary

Problems solved by technology

The disadvantage of this type of algorithm is that as the dimension of the data set increases, the performance of these tree-structured algorithms gradually deteriorates. The reason is that high-dimensional data sets will increase the complexity of the tree structure, which will lead to the establishment of tree structures and search for neighbors. Points and increased time taken to calculate distances

The kNN algorithm is a commonly used data mining algorithm, but its time complexity is high and the classification speed is slow.

Images