File recommendation method and apparatus

Abstract

The invention discloses a file recommendation method and apparatus, and belongs to the field of network technologies. The method comprises: for each user in multiple users, according to multiple file identifiers comprised by historical operation records of the user, and weights of the multiple file identifiers, acquiring multiple key words corresponding to the multiple file identifiers and the weights of the multiple key words; calculating transition probability between each two key words; calculating target weights of the multiple key words; and performing recommendation according to the multiple key words corresponding to a current user and the target weights of the multiple key words. According to the method and apparatus provided by the invention, a recommendation method based on the key words is adopted, user interests are differentiated according to the key words of the files, thereby realizing customization for users and improving the success rate of recommendation. Furthermore, the key word of the file is not affected by file publication time or rating performed by a user on the file, thereby improving the success rate in recommending a new file.

Description

technical field [0001] The present invention relates to the field of network technology, in particular to a file recommendation method and device. Background technique [0002] In daily online activities, users are faced with all kinds of information all the time, but it is difficult to filter out the information that they are really interested in. In order to facilitate user screening, the server may recommend information that may be of interest to the current user. [0003] Taking video as an example, the server can divide multiple users into multiple groups according to basic information such as gender and age. When recommending videos for the current user, first determine the group to which the current user belongs, and then determine the The most watched videos by users are recommended to current users. Alternatively, the server may also use a collaborative filtering algorithm to calculate the similarity of each two videos based on the user's ratings for different vid...

AI Technical Summary

Problems solved by technology

[0004] However, the basic information on which the server divides the groups is difficult to reflect the interests of the users, and the divided groups cannot be personalized. Low; and when the server adopts the collaborative filtering algorithm for recommendation, for new videos, since users rarely watch new videos and score new videos, when calculating the similarity between new videos and other videos, similar The accuracy of the degree is poor, resulting in a low success rate of recommending new videos

Images