Theoretical Research On Cache Optimization Based On Machine Learning In Edge Network Environment

With the development of science and technology,various services and applications are connected to the Internet.The data in the network has experienced a blowout growth,which is dominated by video traffic.Edge network refers to the edge of public network.It makes the delay reduced and the load pressure of the core network offloaded.Edge cache is one of the key technologies.It caches part of resources on the edge server.When the services come again,they can obtain the resources from the cache,and there is no need to obtain the resources from the core network.This thesis needs to analyze a large amount of data.Machine learning has a great advantage in numerical analysis.Therefore,machine learning is used in this thesis.The VOD service is token as an example.A cache optimization and a cache replacement scheme based on machine learning in edge network environment are proposed.The delay is minimized and the load of the core network is reduced.This thesis has high theoretical research value and practical application prospects.The main work of this thesis is as follows:Firstly,the background and significance of this thesis are introduced.Some research and status of cache in the country and abroad is expounded.The organization structure and main work of this thesis are summarized.Secondly,the development of machine learning is described briefly.The application and several classic algorithms of machine learning are introduced.The algorithms(XGBoost algorithm and Random Forest algorithm)that used in this thesis are introduced.Thirdly,the volume of access is predicted and a cache optimization algorithm of VOD service based on XGBoost algorithm is proposed.The information of samples is preprocessed according to the weekly model and daily model.The importance of features is displayed and some features of lower importance are eliminated.The important parameters in the XGBoost algorithm are simulated to narrow the selection range.Then the prediction model is established and the volume of access is predicted.A new factor is proposed which is named cache cost-effective.Then an optimization model is established.The knapsack algorithm is used to solve the problem and a fast solution is proposed.The simulation proves the accuracy of the prediction and the effectiveness of the optimization algorithm.Fourthly,the visit duration is predicted and a cache replacement algorithm of VOD service based on random forest is proposed.The information of videos is preprocessed according to the weekly model and the daily model.The importance of features is displayed and some features of lower importance are eliminated.The important parameters in the random forest algorithm are simulated to narrow the selection range.Then the sample data is modeled and predicted.A new factor is proposed which is named replacement cost-effective.Then a replacement model is established.The implicit enumeration method is used to solve the problem.A new filter condition is proposed which named single video replacement.The simulation proves the accuracy of the prediction and the effectiveness of the optimization algorithm.Finally,the research work of this thesis is summarized.The future research direction is put forward.