<![CDATA[Aiming at the problem of poor feature expression ability and model representation effect of traditional video recommendation mechanism, combined with the characteristics of traditional recommendation algorithm, this paper deeply studies the short video publishing algorithm and recommendation mechanism under artificial intelligence, and constructs a two-layer feature representation model BIFR based on attention. Firstly, the basic principle of recommendation algorithm is introduced in detail, and then the internal representation of features is studied through a multi head self attention mechanism to deeply mine the correlation between features and further improve the expressiveness of features. Then adjust the input feature crossover to learn the feature crossover more effectively. Finally, combine the two, add DNN to get the final output results, and then use the corresponding evaluation indicators to test the constructed recommendation model. The test results show that the video recommendation model constructed in this study has high accuracy, strong expressiveness and effectiveness.]]>
Copyrights © 2022
