融合协同过滤和相遇概率预测的DTN路由算法  

Routing Algorithm Combining Collaborative Filtering and Encounter Probability Prediction in DTN

作　　者：崔建群[1] 晏晖然 常亚楠 高梦楠 马致远 CUI Jianqun;YAN Huiran;CHANG Yanan;GAO Mengnan;MA Zhiyuan(Hubei Provincial Key Laboratory of Artificial Intelligence and Smart Learning,Central China Normal University,Wuhan 430079,China;School of Computer Science,Central China Normal University,Wuhan 430079,China)

机构地区：[1]华中师范大学计算机学院,武汉430079 [2]华中师范大学人工智能与智慧学习湖北省重点实验室,武汉430079

出　　处：《小型微型计算机系统》2025年第3期735-743,共9页Journal of Chinese Computer Systems

基　　金：国家自然科学基金面上项目(62272189,62372206)资助.

摘　　要：在延迟容忍网络(DTN)中,网络的拓扑结构是随机动态变化的,节点之间的间歇性连接使得网络拓扑有很大的不确定性.所以设计一个高效的路由策略,来提高消息成功投递的机率是容迟网络的主要研究方向.本文对喷射等待路由算法进行优化,提出了一种融合协同过滤和相遇概率预测的DTN路由算法CFPP(A DTN Routing Algorithm Combining Collaborative Filtering and Encounter Probability Prediction).该算法关注节点间的社会属性特征,通过协同过滤算法为节点推荐最佳好友集.在消息转发时,优先考虑将消息转发给处于消息目的节点的最佳好友集中的节点.同时提出了基于灰色预测模型的相遇概率预测算法,将节点间的历次相遇概率作为初始序列进行建模,来预测两个节点下一次相遇的概率.仿真实验表明,与SCSS,CPNASW,TLRNSA,SaW,DPN-ASW算法相比,CFPP算法在消息投递率、网络负载率方面均要优于上述算法.In delay tolerant network(DTN),the topology of the network is randomly and dynamically changing,and the intermittent connections between nodes make the network topology highly uncertain,so improving the chance of successful message delivery by selecting excellent relay nodes for message to forwarding is the main research direction of DTN.In this paper,we optimize the sprayandwait algorithm and propose a DTN routing algorithm combining collaborative filtering and encounter probability prediction(CFPP).This algorithm focuses on the social attribute characteristics between nodes,and recommending the best friend set for nodes through collaborative filtering algorithm.When forwarding messages,priority should be given to forwarding them to the node in the best friend set of the destination node.Meanwhile,a grey prediction model based algorithm for predicting the probability of encountering each other was proposed,which models the probability of encountering each other as an initial sequence to predict the probability of the next encounter between two nodes.Simulation experiments show that the CFPP algorithm outperforms other routing algorithms in terms of message delivery rate,network load factor when compared to SCSS,CPNASW,TLRNSA,SaW,and DPN-ASW algorithms.

关 键 词：协同过滤 延迟容忍网络 灰色模型 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]