Relationship Awareness And Resource Allocation In Fog Computing Based Radio Access Networks

With the continuous evolution of wireless network technology,the rapid popularization of intelligent terminals and the explosive growth of the use of data services and multimedia services,the communication network has become more and more closely connected with social networks,attracting many researchers' attention and opening a new perspective.Among them,social relationships are the most basic social characteristics,and they describe the strength of the connection between two interconnected individuals.Social relationship awareness is a method of inferring social relationships indirectly.In this paper,the methods of social relationship awareness and resource allocation are studied in fog computing based radio access networks.First,based on mobile phone users' call data,short message data and location data,the machine learning method is used to aware the social relationship between users.Then,social relationships are introduced into the problem of resource allocation to optimize the performance of the network.The main research work and innovation points can be summarized as the following aspects:1.Aiming at the problem that the supervised learning method in machine learning has low prediction accuracy when a small amount of social relationships tagging data is used,Co-Forest algorithm based on the semi-supervised ensemble learning is used to aware the social relationships.At MIT's Reality Mining dataset,Co-Forest algorithm achieves an accuracy of 89%with a social relationship label of 5%,and there are 24%and 2%of the increase compared with Support Vector Machine of supervised learning and Co-Training algorithm of semi-supervised learning respectively.The accuracy fluctuates within 3%with different label data percentages.By Co-Training algorithm,the accuracy and robustness of the social relationship awareness method is improved under a few social relationships tagging data.2.Aiming at the problem that centralized server has a heavy computational load and the fronthaul has a heavy burden because large amount of raw data is uploaded,this paper proposes an edge-layered Co-Forest algorithm based on the semi-supervised ensemble learning.By using the computing power of the fog computing node at the edge of the network,it reduces the burden on the centralized server and the fronthaul,and multiple fog computing nodes compute in parallel to reduce the time overhead.3.Aiming at the problem that the security and users' self-confinement limit the application of Device-to-Device(D2D)communications fog computing based radio access networks,social-aware D2D is introduced to allocate resources.A non-convex optimization problem of joint content downloading mode selection,beamforming design,and power control is formulated,and this problem is solved by a branch and bound and weighted minimum mean square error based algorithm.Simulation results show that social-aware D2D communications can improve the total network throughput and effectively offload the throughput of the fronthaul.This paper proposes a social relationship awareness method that is suitable for practical communication networks and uses the aware social relationships to optimize the performance of communication networks.It provides a way of thinking for the combination of social networks and communication networks.