Object recommendation method and device and related equipment

Abstract

The embodiment of the invention discloses an object recommendation method and device and related equipment. The method comprises: constructing a homogeneous user network according to a social relationship between users; establishing a heterogeneous network according to the user, the preference object, the preference relationship between the users and the preference object and the social relationship between the users; performing random walk according to a preset meta-path to obtain a node sequence, and converting the node sequence into a word vector sequence; traversing the homogeneous user network according to the target user; obtaining a social user having a social relationship with the target user; obtaining a word vector corresponding to the target user and a word vector correspondingto the social user according to the word vector sequence searched by the target user and the social user; and obtaining user similarity between the social user and the target user according to the word vector corresponding to the target user and the word vector corresponding to the social user, and finally recommending a preference object corresponding to the social user to the target user according to the user similarity, thereby realizing object recommendation to the target user.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to an object recommendation method, device and related equipment. Background technique [0002] With the rapid development of Internet technology, there are more and more scenarios for actively recommending objects to users. For example, music software recommends music works to users, and e-book reading platforms recommend e-books to users. Object recommendation is directly related to user experience. [0003] Among the existing object recommendation methods, there are object recommendation methods that use graph neural networks to process graph networks to obtain recommended objects, and methods that use relationship chains for object recommendation; where the object is a song as an example, according to the user's relationship network, The user's preferred song and the singer corresponding to the song establish a corresponding graph network (the graph network includes t...

AI Technical Summary

Problems solved by technology

[0003] Among the existing object recommendation methods, there are object recommendation methods that use graph neural networks to process graph networks to obtain recommended objects, and methods that use relationship chains for object recommendation; where the object is a song as an example, according to the user's relationship network, The user's preferred song and the singer corresponding to the song establish a corresponding graph network (the graph network includes three types of nodes: user, song, and singer). After using the existing graph neural network to perform random walks on the graph network, many infinite Meaningful point sequence, such as user A->user B->song C->singer D, if this type of point sequence is continued to perform object recommendation calculations, the calculation accuracy of the recommended object will decrease, and the deviation of the object recommendation result is relatively large. Large, resulting in a decrease in the accuracy of the recommendation

However, using relationship chains for recommendation, the objects (such as songs, e-books, or commodities) that the user’s friends are interested in are directly recommended to the user. The objects of interest of the friends of the user are regarded as the objects of interest of the user for object recommendation. When some of the friends of the user have a large deviation from the interest of the user, the result of the object recommendation will have a large deviation, resulting in a decrease in the accuracy of the recommendation.

[0004] It can be seen from the above that in the existing object recommendation methods, due to the meaningless point sequence after the random walk, the object recommendation method based on the relationship chain cannot solve the problem of friend interest deviation, which leads to low accuracy of object recommendation. poor experience

Images