Caching scheme based on user clustering and user requests prediction in small cells

In order to cope with the relentless data tsunami brought by the growth of mobile video service, this paper proposes a two-stage caching scheme. Considering the characteristic of the ultra-dense network that many users may be in the overlapping coverage of multiple small cells, we cluster those users with k-medoids clustering algorithm and the users of the same cluster access to the same matching small cell in the first phase. In the second phase, a cache algorithm is proposed to make cache decision of video files on the basis of predicting the future requests of the users in the same small cell by collaborative filtering algorithm. At last, the simulation results show that the gain of the proposed caching scheme over the existing caching schemes in the cache hit rate and the average delay.