基于评分网络的含时信誉评价算法研究

发布时间：2018-04-08 12:32

本文选题：社交网络　切入点：时间相关　出处：《深圳大学》2017年硕士论文

【摘要】：随着在线社交网络的快速发展,人们越来越重视信息的共享,基于Web的评分系统在电子商务与消费点评网站中的应用也越来越广泛。此类应用中存在着用户和对象两种实体。用户可以给对象进行评分,系统根据用户的评分计算出对象的最终评分,而最终评分的高低影响着对象的排名。因此如何提高在线评分系统的准确度是近年来数据挖掘领域的研究热点。现有较多研究主要通过引入用户信誉度的概念,来衡量用户信誉度的高低来重新调整用户打分的权重,从而提高排名算法的健壮性。针对如何提高在线评分系统准确度的问题,本文对电子商务环境下的信誉机制进行了研究,并主要提出了一个含时信誉评价算法,研究的主要内容如下:(1)研究了现有的信誉模型,分析各种模型的信誉值计算方法,掌握评估信誉的思路,总结了影响信誉值计算的各项因素,以及对信誉模型的攻击和抵御的方法。同时也分析了电子商务中在线信誉系统的应用现状,以及电子商务环境下信誉系统的发展方向,为后续的论文研究奠定了理论基础。(2)通过研究社交网络中影响用户信誉和对象质量的因素,提出一个含时信誉评价算法。该算法主要通过从时间阶段性角度和数量角度引入了行为权重因子、信誉积累过程和质量积累过程来重新定义用户的信誉和对象的质量。此外,为了更加深入的掌握信誉的影响能力,研究并分析了相关信誉模型,在借鉴其信誉排名模型的基础上提出了一个适用于电子商务环境下的信誉模型。本文给出了优化后的模型具体定义、设计思路和程序流程,同时在理论上做了验证,并通过设计实验来验证了模型的有效性。(3)在多个真实的数据集上面对含时信誉评价算法进行验证。通过使用三个数据集与相关算法进行了实验测试,采用准确率、AUC、召回率和F函数等几个指标进行对比。结果表明采用这个算法的信誉值更优越且稳定,并且所获得的质量排名结果有很大的提高,从而验证本算法的有效性和可行性。
[Abstract]:With the rapid development of online social networks, people pay more and more attention to the sharing of information, and the application of Web based scoring system in e-commerce and consumer comment sites is becoming more and more extensive.There are two kinds of entities in such applications: user and object.The user can grade the object, and the system calculates the final score according to the user's score, and the final score affects the ranking of the object.Therefore, how to improve the accuracy of online scoring system is a hot topic in the field of data mining in recent years.In order to improve the robustness of ranking algorithm, more researches have been done mainly by introducing the concept of user reputation to measure the level of user reputation to readjust the weight of user rating.Aiming at the problem of how to improve the accuracy of online scoring system, this paper studies the reputation mechanism in electronic commerce environment, and puts forward a time-dependent reputation evaluation algorithm.The main contents of the study are as follows: (1) the existing reputation models are studied, the credit value calculation methods of various models are analyzed, the train of thought of evaluating reputation is grasped, and the factors influencing the credit value calculation are summarized.And the attack on the reputation model and the method of defense.At the same time, it also analyzes the application status of online reputation system in electronic commerce, and the development direction of credit system in electronic commerce environment.A time-dependent reputation evaluation algorithm is proposed by studying the factors affecting user reputation and object quality in social networks.In addition, in order to grasp the influence of reputation more deeply, this paper studies and analyzes the relevant reputation models, and puts forward a reputation model suitable for e-commerce environment on the basis of its reputation ranking model.In this paper, the specific definition of the optimized model, the design idea and the program flow are given, and the theoretical verification is given.The validity of the model is verified by designing experiments.By using three data sets and the related algorithms for experimental test, the accuracy of AUC, recall rate and F function are compared.The results show that the credibility value of the proposed algorithm is superior and stable, and the quality ranking results obtained are greatly improved, which verifies the effectiveness and feasibility of the proposed algorithm.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09;TP311.13

【相似文献】