A high-accurate content popularity prediction computational modeling for mobile edge computing using matrix completion technology

Mobile edge computing (MEC) is envisioned as a promising platform for supporting emerging computation-intensive applications on capacity and resource constrained mobile devices (MDs). In this platform, the task with high computing resource demand can be offloaded to edge nodes for computing. Moreover, the computing result can be cached to edge nodes. When other MDs request the task that has been cached, the edge nodes can directly return the result to MD. However, the storage capacity of edge nodes is limited, the effective task prediction and caching scheme is one of the key issues for MEC. In this article, a matrix completion technology based content popularity prediction joint cache placement (MCTCPP-CP) scheme is proposed to tackle this issue for MEC. On the one hand, the MCTCPP-CP scheme is the first scheme using matrix completion (MC) technology to content popularity prediction. It proved by experiments that the accuracy of using MC technology to estimate caching content is improved compared with the previous methods. On the other hand, a cache placement decision approach based on the benefit of unit storage is proposed. Extensive numerical studies demonstrate the superior performance of our MCTCPP-CP scheme. The key performance indicators such as task duration, hit rate, estimated error are better than previous schemes by about: 0.13% to 14.01%, 17.28% to 37.65%, and 8.17%.