Method and system for recommending videos

Abstract

A method for recommending videos is provided, wherein the method comprises following steps: obtaining video operation history information of a user; predicting a video recommendation list of the userby using a recessive feature algorithm according to the video operation history information; recommending the video recommendation list to the user. The embodiment of the invention can solve the feature combination problem under the dilution data, mathematically quantify the implicit relationship between the features, more accurately predict the possibility of the user clicking the video, and improve the accuracy of recommending the video for the user.

Description

technical field [0001] The present invention relates to the technical field of video analysis, in particular to a video recommendation method and system. Background technique [0002] At present, most video recommendation strategies are implemented based on collaborative filtering or labeling systems. The collaborative filtering algorithm calculates relevance and gets recommendations based on user ratings. Collaborative filtering algorithms are based on shared views among users. In other words, the collaborative filtering algorithm is based on the assumption that the way to find the content that a user is really interested in is to first find users with similar interests to him, and then recommend the content that this type of user is interested in to this user. The labeling system uses experts to label and classify products, and then recommend products with the same label as the user's history. [0003] However, there are various problems in collaborative filtering algor...

AI Technical Summary

Problems solved by technology

Three, the new user / item (NewUser / Item) problem, for the new user problem, since there is no score for the item, the similarity cannot be calculated, and of course no recommendation can be generated

For new projects, similar problems exist, the system recommends poor quality at the beginning

Fourth, scalability (Scalability) problem: based on the nearest neighbor algorithm, the increase in the dimension of items and users will lead to a very large amount of calculation, so it is very important to consider the scalability of the algorithm

The volume of videos is very large, and there are also a lot of new videos generated every day. It is a long process to tag only by manual

Second, the label granularity problem, the granularity of the label is difficult to determine

Third, the subjective influence

It is difficult to guarantee the consistency of labeling by these experts

Images