Content Centric Network With Label Aided User Modeling and Cellular Partition

The popularity of smart devices and the increase of access rate are switching Internet from link-centric (host-to-host) to content-centric (user-to-content). Since the primary purpose of users is to find their desired contents from the Internet, the traditional host-to-host structure will cause lots of useless traffic and decrease the network efficiency. Content-centric network (CCN) is then introduced to solve this embarrassed situation. As a new-born technology, CCN has not fully explored its potential to utilize the knowledge embedded in the contents. Hence, we propose to enhance the information extraction capability of CCN with data driving methods. Specifically, we add a label vector containing feature descriptions for each content, and then apply neural network (NN) to model users' behavior and predict their responses to new contents. To further increase prediction accuracy and potential benefits, we group users into different cells. The behavior sets of users in a cell at different time intervals are modeled by a series of NNs. Moreover, we design a new caching policy to improve network caching efficiency. Benefits in other aspects and future research directions are also discussed. Quantitative evaluations of the proposed model and caching policy are presented in the end of this paper.