Research On Popularity-Driven Video Caching Strategies

Deploying caching in base stations can effectively reduce the transmission delay of video and avoid backhaul link congestion.However,the base station has a small cache capacity,and it is impossible to cache all video in the base station.Caching popular content is an effective way to solve this problem.Therefore,how to design an efficient popularity prediction algorithm to predict the popularity of video and how to formulate an effective collaborative caching strategy are key issues that base station caching needs to solve.This paper mainly studies video popularity prediction and collaborative caching strategies in mobile networks.First,the method based on deep learning is used to predict the popularity of video,and a popularity prediction model that combines content features and temporal process is proposed;Then the video's popularity is sorted,and a cooperative caching strategy based on transmission delay is further designed.following is the main work of the paper:(1)The popularity prediction model based on deep learning is studied,and a deep attention video popularity prediction model(DAFCT)that combines content features and time temporral is proposed.First,build a long and short-term memory network(Attention-LSTM)based on the attention mechanism to capture the growing trend of video popularity.For high-dimensional multi-modal content features,embedding technology is used to reduce the dimensionality,and the embedding vector of the content feature is input into the second-order interactive pool for feature combination.Then,the concatenate method is used to fuse the popular trend and content features,and the multi-layer perceptron is used to learn the fused features to predict the popularity of the video,and a Deep Attention Video Popularity Prediction Model(DAFCT)that combines content features and time temporal is proposed.Finally,a Deep Attention Video Popularity Prediction(DAVPP)algorithm is designed to verify the effectiveness of the DAFCT model.Experiments show that compared with AttentionLSTM and NFM models,the recall rate is increased by 10.82% and 3.31%;the F1 score is increased by 9.80% and 3.07%.(2)The cooperative caching strategy(HCP)is studied,and a cooperative caching strategy(CCSTD)based on transmission delay is proposed.HCP is a collaborative caching strategy developed under the assumption that the popularity of the video is known,and it has limitations.In order to further improve the user experience,the popularity ranking method in the HCP strategy is first used to sort the video predicted by the DAVPP algorithm according to the popularity from high to low.Then the sorted video is allocated to the base station cache,the CCSTD strategy is proposed,and the delay of user retrieval of the content is analyzed.On this basis,the CCTDA algorithm is designed to verify the effectiveness of the CCSTD strategy.Finally,the simulation experiment of HCP and CCTDA algorithm is carried out.Experimental results show that the video transmission delay of the CCTDA algorithm is lower than that of the HCP algorithm,and the cache hit rate is higher than that of the HCP algorithm.