Research On Random Forest-based Socially-aware Routing Strategy

With the popularity of the smart mobile devices and the development of the wireless technologies,humans who carry these devices can be connected ubiquitously.Because of the mobility of the devices controlled by individuals,it is a rough task to maintain the end-to-end path between source node and destination node.Researchers have introduced the analysis of nodes’ social behavior to solve the problem of data dissemination in the networks,which leads to the emersion of the Mobile Social Networks.Due to the lack of network infrastructures,data is mainly disseminated in MSNs by means of a carry-and-forward mechanism.Various literatures have been proposed to promote more efficient data forwarding performance and provide humanized service by introducing social features and digging social properties from the networks.Recently,some mechanisms that apply dynamic social features have been proposed to enhance the efficiency of the message transmission.Thesis validates that dynamic social features can result in changes of communities which are divided in terms of these features.And the traditional community division methods are mainly based on the static social features and cannot be directly applied to the situation that nodes’ social features change dynamically.Therefore,this paper uses the random forest classifier in machine learning to dynamically divide the nodes in networks.The k-means++ algorithm is used to generate a training set for the classifier.In order to measure the quality of the generated training set,thesis designs a contact coefficient CR.Based on the trained random forest model,a Random Forest-based Socially-Aware Routing Strategy(SARP)is proposed.To the best of our knowledge,this is the first method that combines dynamic social features with dynamic community classification.In the simulation experiment,the proposed algorithm is compared with the classic Flooding,SANE and Multi-Sosim.The experimental results show that our routing protocol can effectively improve the data transmission ratio and reduce the transmission latency in networks.This proves the superiority and feasibility of the algorithm in Mobile Social Networks.