Video cold start recommendation method and system

Abstract

The invention discloses a video cold start recommendation method and system. The video cold start recommendation method comprises the steps: S1, carrying out the dimension reduction of a new video based on an Inception network, and generating a video vector for the new video; S2, storing the video vector in Faiss; S3, summing corresponding video vectors by adopting five videos recently watched bya user, taking a mean value as a user vector, and indexing Faiss; and S4, returning a video corresponding to the video vector with a small distance from the user vector to the user. According to the video cold start recommendation method, the new video is subjected to frame capture processing to form the plurality of pictures, and the feature vector is generated for each picture to generate the video vector, and the vector index is carried out based on the Faiss to carry out video recommendation, so that the cold start recommendation of the video is realized, and the complexity is low, and therecommendation efficiency is high.

Description

technical field [0001] The present invention relates to the field of content recommendation, in particular to a video cold start recommendation method and system. Background technique [0002] With the popularity of various applications, enterprises can collect more and more complete user data. How to use these data to increase revenue is a problem that all enterprises will face. The most common way is personalized recommendation, especially in e-commerce, video sites or other content platforms. The main goal of personalized recommendation is to recommend a large number of objects to a large number of users who may like it, such as recommending videos of interest to users. [0003] For any Internet content platform, a large number of objects and users are constantly growing and changing. The cold start of the recommendation system refers to how to recommend objects to new users for newly registered users or newly entered objects. Satisfied, how to distribute the new subjec...

AI Technical Summary

Problems solved by technology

[0005] However, in tag-based recommendation, the accuracy and recall of machine learning prediction tags are not enough, especially for some low-quality videos, the accuracy and recall of tag-based recommendation are extremely low

In addition, the cost of manual labeling is too high, and the volume of hundreds of thousands of videos released every day is still increasing

In the recommendation of content information based on video text and other content information, the text information of the video is too little, and many videos are released without text information. Therefore, it is difficult to construct a vector for the video; in addition, the traffic of text word segmentation is slow and time-consuming. low efficiency

Images