A fast approximate k-nearest neighbor method based on tree strategy and balanced k-means clustering

Abstract

The invention provides a fast approximating K nearest neighbor method based on tree strategy and balanced K-means clustering, so as to improve the performance and speed of approximate K neighbor retrieval. First of all, the balanced K-means tree is constructed by the balanced K-means clustering method, so that the data can be efficiently and orderly organized in a tree structure, and the rapid positioning of any new sample data can be realized; then, using the anchor location method and the idea of ​​adjacent clusters, Quickly find multiple approximate neighbors of new data samples through the balanced tree, that is, K nearest neighbors. The method of the invention takes into account the advantages of the tree-based K-nearest neighbor algorithm and the balanced K-means algorithm, and can be applied to multiple fields such as image recognition, data compression, pattern recognition and classification, machine learning, document retrieval systems, statistics and data analysis, and the like.

Description

technical field [0001] The invention belongs to the technical field of machine learning and data mining, and in particular relates to a fast approximate K nearest neighbor method based on tree strategy and balanced K-means clustering. Background technique [0002] In the era of mobile Internet, people's daily life is faced with the impact of massive data every day, such as personal information, video records, image collection, geographic information, log files, etc., in the face of such a large and growing data information, how to process all The effective storage, indexing and querying of the required information is a hot research topic both at home and abroad. [0003] Approximate K-nearest neighbor retrieval was initially applied to document retrieval systems as a method for finding similar document information, and then in geographic information systems, K-nearest neighbor retrieval was also widely used in location information, query, analysis and statistics of spatial d...

AI Technical Summary

Problems solved by technology

For example, K-nearest neighbor retrieval in large-scale high-dimensional data has always been one of the hot issues that are difficult to overcome. In the tree structure method, KD tree, KD random forest, etc. have good results, but in the KD tree algorithm, The retrieval process needs to go back to the previous node continuously. The higher the dimension, the more backtracking times are required, and the efficiency of the algorithm will be lower. In KD random forest, although the backtracking problem can be alleviated to a certain extent, due to KD Random forest uses multiple KD trees to search together. How to balance memory usage and algorithm efficiency has become a new problem

Images