Research On Content Popularity Prediction And Caching Strategy Of Cache Enabled Celluar Networks

With the continuous rise of mobile Internet technology and the expansion of the production and application scale of smart mobile devices,various mobile applications emerge in an endless stream,and the traffic volume generated by mobile networks shows a rapid growth beyond expectations.How to not only reduce the huge load pressure faced by the system backhaul link,but also meet the user’s requirements for high-quality communication services is a technical problem that needs to be solved urgently in the current mobile network.In order to effectively solve the above problems,the current mainstream trend considers introducing edge caching technology into the traditional cellular network,and caching files according to interest of users in advance in cellular network-related edge devices with certain storage capabilities.At the same time,since the storage capacity of the base station is limited by physical conditions,it is necessary to formulate a reasonable cache placement and content transmission strategy that fits the system characteristics,so as to reduce the load pressure on the cellular networks brought by the mobile traffic transmission process.The topic of thesis is selected from the National Science Foundation Project "Research on Multi-level Cooperative Caching Method in UAVassisted Cellular Networks".Focusing on the dynamic changes of content popularity and the limited storage capacity of base stations,thesis proposes a content popularity prediction method suitable for caching cellular networks,and proposes a caching strategy under cooperative caching of base stations according to the content popularity prediction results.The main work of the thesis is as follows:1)The edge caching technology applied in the cache enabled cellular networks is reviewed.Firstly,the characteristics of edge caching technology and the specific application of this technology in actual situations are described.Secondly,several common types of edge caching technologies and the advantages of edge caching are summarized.technical characteristics.Thesis discusses the current hot issues in edge caching technology,and summarizes the content popularity prediction technology and commonly used caching strategies.In the review of content popularity prediction,the key problems existing in the current content popularity prediction research are analyzed,and the existing prediction methods are summarized and classified.In the review of caching strategy research,the current mainstream caching strategies are summarized and classified,and the commonly used performance evaluation index system optimization goals in mainstream caching strategies are analyzed and summarized.2)In the cache enabled cellular network scenario,aiming at the problem that the content popularity prediction in the existing research ignores the spatial correlation in user requests,a content popularity prediction algorithm based on spatial-temporal two-dimensional features is proposed.The proposed algorithm uses the current popular neural network prediction method to construct the cached cellular network as a network topology,uses the graph neural network to capture the spatial correlation between base station nodes,and uses the time series prediction network to capture the temporal correlation of the base stations content request sequence,Finally,features that contain spatial-temporal correlations are obtained.The prediction model is trained according to the extracted features,and the data to be tested is input into the trained model,and the spatial-temporal information in the data can be used to realize the prediction of the popularity of the content.Simulation results show that,compared with several classical algorithms commonly used in current research,the proposed algorithm can improve the prediction accuracy of content popularity by at least 3%.3)In the cache enabled cellular networks,in order to solve the problem of dynamic changes in content popularity and limited storage capacity of base station caching devices in actual situations,an optimization algorithm for caching placement and content transmission is proposed.Firstly,the system cost brought by caching cellular networks response to user requests is defined,and the system cost minimization problem is constructed,with the purpose of reducing the system transmission cost as much as possible and optimizing the local cache hit rate.Thesis divides the caching strategy into two processes:cache placement and content transfer.For the cache content placement strategy,thesis uses the TGCN algorithm to predict the content popularity at the base station node,and the base station places the cache content according to the prediction results.For the cache content transmission strategy,thesis uses Lagrangian relaxation and the method of primal dual problem decomposition to decompose the coupling variables,and solve the variables separately.For the original problem including coupling variables,thesis decomposes it into two sub-problems,one sub-problem mainly solves how to realize the content transmission of the access base station,and the other sub-problem mainly solves how to realize the content transmission of the cooperative base station,these two variables are in the main problem are mutually coupled.When solving the content transmission variable of the access base station,first use the one-to-many matching algorithm to obtain the user access strategy.If the user request hits the local cache file of the access base station,the access base station will transmit the cached content to the user.For the user’s unresponsive request,the weight of the file is calculated,and the coordinated base station selects the file for content transmission according to the weight.The variable is brought into the optimization problem and the dual problem to iteratively solve,and finally the optimization of cache placement and content transmission is realized.Thesis uses the PYTHON and MATLAB platforms to simulate the caching cellular network communication scenario and the proposed algorithm,and compares it with other algorithms to verify the convergence and effectiveness of the algorithm.According to the data obtained from the simulation,the algorithm proposed in thesis can effectively reduce the system cost,optimize the local cache hit rate,and reduce the transmission burden of the backhaul link.