Method and system for recommending object to terminal user

Abstract

The invention provides a method for recommending an object to a terminal user. The method comprises the steps of based on N objects, establishing an object correlation data set, wherein the object correlation data set comprises all objects in the N objects browsed by any terminal user within a first time period, and a correlation group consisting of two objects in the N objects successively browsed by any terminal user within the first time period according to a browsing sequence; determining the individual frequency of each object in the object correlation data set, occurring in the object correlation data set and the correlation group frequency of the object correlation group occurring in the object correlation data set, calculating the correlation degree of the two objects in the object correlation group based on the individual frequency of each object in the object correlation group and the correlation group frequency, and determining the object calibration recommendation degree for a terminal user in the objects based on the correlation degree and the determined seed object.

Description

technical field [0001] The present invention relates to network data processing, and more particularly to a technology for recommending objects to end users. Background technique [0002] The recommendation system is the most typical and important application in the field of data mining in the Internet industry, e-commerce, and retail industries. [0003] At present, the recommendation systems of most companies are implemented based on collaborative filtering algorithms. Although collaborative filtering algorithms have shown good performance in application and product recommendation, there is still room for improvement. The collaborative filtering algorithm builds user evaluations based on the consumption similarity between users to make recommendations, and the similarity between users is identified by the similarity between high-dimensional sparse vectors. The obtained similarity between users may be inaccurate, resulting in inaccurate recommendation results, and as the n...

AI Technical Summary

Problems solved by technology

The collaborative filtering algorithm builds user evaluations based on the consumption similarity between users to make recommendations, and the similarity between users is identified by the similarity between high-dimensional sparse vectors. The obtained similarity between users may be inaccurate, resulting in inaccurate recommendation results, and as the number of merchants increases, the degree of vector sparsity calculated by the recommendation system will increase, which not only reduces the accuracy of recommendation At the same time, it increases the computational complexity and reduces the performance of the recommendation system

Images