Discrete multi-view hash-based collaborative filtering recommendation method

Abstract

The invention discloses a discrete multi-view hash-based collaborative filtering recommendation method. The method comprises the following steps of 1) constructing a multi-view anchor point graph representation of data according to the data in different views; 2) in combination with collaborative filtering and an anchor point graph, obtaining a learning model; 3) solving the obtained learning model to obtain binary hash codes, corresponding to articles, of a user; and 4) performing a nearest neighbor search by utilizing the obtained hash codes, calculating preference degrees of a specific user to candidate articles, and returning a plurality of the articles with the maximum preference degrees to serve as recommendation results. According to the method, the data in the different views is integrated, so that the discrete characteristics of the codes are always kept during solving and the quality of the recommendation results is improved; and meanwhile, quick search of similar users is realized by utilizing the hash codes, so that the recommendation result calculation efficiency is improved.

Description

technical field [0001] The invention relates to a personalized recommendation technology, in particular to a collaborative filtering recommendation method based on discrete multi-view hashing. Background technique [0002] The rapid development of the Internet industry has brought explosive growth in content. In order to help users obtain information effectively, personalized recommendation systems are playing an increasingly important role. Collaborative filtering technology is a kind of technology that has received wide attention in recommender systems. Different from the traditional recommendation based on content-based direct filtering analysis, collaborative filtering uses a large amount of user information to select users similar to the target user or items similar to the target item to finally recommend items that the current target user may be interested in. [0003] But in the real application environment, we can often obtain a lot of other information besides the...

AI Technical Summary

Problems solved by technology

Traditional collaborative filtering recommendation technology can only use user information in a single view, and needs to calculate the prediction score of user preference through the operation between high-dimensional vectors, which seriously affects the calculation and storage efficiency. The conventional solution Also caused a lot of information loss

Images