Content Popularity Prediction with Integrated Features in UAV-Assisted Wireless Caching Networks

Abstract

Mobile content service has been experiencing an explosive traffic growth in radio access networks. Due to the frequent downloads of popular content requested by multiple mobile users, most of the data traffic is caused by repetitive data transfers. Proactive content caching is an effective way to reduce the burden of backhaul link traffic and to improve the user experience. Content popularity is an important factor which influences active caching. However, content popularity is often unknown in advance. Therefore, predicting content popularity is an important challenge for content caching. It is studied that the problem of cache content popularity prediction in a unmanned aerial vehicle(UAV) assisted wireless cache network(UA-WCN) which includes a base station(BS) and multiple UAVs. A model framework and strategy for content popularity prediction based on the comprehensive characteristics of users and content are proposed to solve the popularity prediction problem. Firstly, user preference learning is applied based on considering that users prefer request the contents they are interested in. The users is classified by using some of the physiological characteristics of the user. Then, the weight of the user's preference for the content is learned by using Follow The Regularized Leader(FTRL) algorithm. Finally, the gradient descent algorithm is used to learn the local model parameters. The feasibility of the model is verified by the simulation results. And the proposed popularity prediction strategy can not only accurately predict the popularity of content, but also greatly reduce the computational complexity.