【作者】 熊建英；

【导师】 钟元生；

【作者基本信息】 江西财经大学 ， 管理科学与工程， 2013， 博士

【摘要】 C2C作为传统电子商务中最活跃的一种模式,与B2B和B2C模式相比,除了虚拟性与时空跨越性外,还包括个体参与、市场进出壁垒低、参与者可以匿名注册等特性,这些都会造成交易双方信息的不对称,导致交易不确定性,增加消费者风险感知。而随着移动商务的发展,基于位置、情境的服务技术正深刻的影响着电子商务产业的变革,推动C2C模式的改革创新。情境(上下文)相关的C2C模式给用户带来巨大商机,正逐渐被人们所接受。这种交易上下文相关性特征,也进一步加剧了信息的不对称性,给欺诈带来更多可趁之机。已有研究与实践表明建立信任关系可以降低交易不确定性和风险性,提高合作效率,抑制交易欺诈。虽然,针对传统C2C的信任管理已取得不少研究成果,但在信誉欺诈处理上还存在一些漏洞,而且缺少对这种上下文相关的C2C交易模式的信任评估研究。因此,分析不同C2C交易场景,建立在线信任管理规避交易风险,从而打造诚信交易环境已成为C2C应用与未来发展的重要问题。本文将从C2C电子商务交易模式出发,分析两种具有代表意义的交易场景,包括：(1)传统C2C交易场景,该场景是目前普遍被人们接受的商务形式,例如淘宝网；(2)上下文相关的C2C交易场景,该场景是在移动商务背景下,随着基于位置服务发展而提出的,C2C平台可以根据上下文进行交易推荐,用户也可以进行交易搜索,然后通过协商,最后进行线下交易的一种方式,如拼车服务。本文在基于这两类应用背景,归纳出研究需要解决的两个主要问题。(1)传统C2C交易信任管理中,需要分析信誉欺诈行为特征,以及已有C2C信誉模型存在的问题,最后进行改进以提高信誉模型的抗欺诈能力。(2)上下文相关的C2C交易信任管理中,由于交易是需要在线下一定交易上下文中完成,不同上下文对交易者会产生不同的风险感知,再加上该商务模式并不成熟,容易出现反馈稀疏、缺少同等上下文下的反馈参考,这使得信任评估需要考虑降低信任的偶然性,并处理不同风险上下文的信誉反馈。针对所要解决的问题,主要研究工作可以细分为以下四个方面。(1)对传统C2C交易场景进行描述,分析已有C2C信任管理中,信誉累计模型所容易滋生的信誉欺诈问题,详细描述了本文所提出的C2CRep模型建立准则、参数选取、各因子计算依据,及实验评价。(2)分析C2C信任管理中买方评论反馈对潜在买家决策的影响作用,提出一种依靠评论反馈可靠度排名方法,将可靠度高的评论优先显示从而提高评论阅读效率。为了计算反馈可靠度,本文改进加权的RFM客户价值模型以适应C2C环境,分析了RFM中各指标计算过程,通过AHP方法以确定RFM各指标权重,最后综合客户价值理论与客户信誉建立反馈可靠度计算模型。(3)描述上下文相关的C2C交易场景,分析不同上下文所产生的风险评估问题。借鉴信息系统风险评估方法,提出一种基于模糊集合理论的方法,分析了风险上下文因素层次结构、各层因素风险模糊定量分析、最终得到风险综合评判结果。(4)针对上下文相关的C2C交易中容易出现反馈稀疏、不同上下文的信任映射问题,本文提出依靠反馈偏移度进行不实反馈过滤的方法。通过反馈数量计算反馈稳固度,结合内容(3)中对上下文风险评估研究,通过风险值进行不同上下文的信任映射,建立一种上下文风险感知C2C信任管理模型。在以上研究内容中,本文所存在的创新点主要包括：(1)提出了C2CRep卖方信誉模型,该模型增加共谋因子平衡了交易次数与反馈过于集中问题；增加了差评度指标抑制反馈敲诈行为；改进价格衡量指标以动态适应商品类型个性差异。与传统C2C模型和SPORAS经典模型相比,实验结果表明,在“低价商品信誉欺诈”,“信誉共谋”,“信誉诋毁”方面都有更强的抵抗作用。(2)提出一种基于RFM模型的买方评论反馈可靠度计算方法,该方法综合客户价值与信誉理论,并改进了RFM客户价值模型以适应C2C电子商务环境。实验表明,与传统利用信誉的方法相比,增加了通过虚假交易获取反馈评价排名的交易时间、频率、金额的欺诈成本。(3)提出一种上下文风险感知的C2C信任模型,该模型基于模糊理论对上下文风险进行评估,通过交易中上下文风险值进行不同情境下的信任映射,同时在基于反馈偏移度的不实反馈过滤中考虑了反馈的稳固度,一定程度上缓解了反馈数量不足导致的偶然信任。仿真结果表明,与传统均值模型、PeerTrust模型相比,提高了信任评估的准确性,抑制了低风险上下文的信任榨取问题。通过本文的研究工作,取得了一点阶段性成果,这将一定程度上丰富电子商务信任管理机制,也将进一步推进未来C2C商务的应用发展。更多还原

【Abstract】 C2C is the most active model in e-commerce. Compared to B2B and B2C, in addition to the virtual and space across, C2C has its own features including:individual participants, low barriers to entry and exit, anonymous registration and so on. It determines an uncertainty, highly dynamic and virtual trading environment which makes the high risk for C2C trading. With the development of mobile e-commerce, Location-based Service and Context-based Service will have a profound impact on the e-commerce, and conduct the innovation of C2C model. Context-sensitive C2C e-commerce will bring more business, and be accepted by user. The context-sensitive feature aggravates the asymmetry of information, and brings more fraud. The related work shows that establish a trust relationship can reduce transaction uncertainty and risk, improve the efficiency of cooperation and restraint of trade fraud. Although trust management for traditional C2C has been made a lot of research, but there are still many problems in handling reputation fraud, and lack of trust study about context-sensitive C2C. So the analysis of different C2C trading scenario is necessary, construct online trust management to avoid trading risks, and create an integrity trading environment has become a key problem in C2C application and future development.The paper analyzes two significance transactions including:(1) traditional C2C scene, it is generally accepted by people, such as Taobao;(2) context-related C2C scene, it is under the development of Location-based Services in mobile e-commerce environment, C2C platform can recommend trading according to the context, or allow users to search trading. And finally users can deal the trading offline after negotiation,, such as carpooling service. In this paper, two main issues need to be considered based on the two C2C scenes.(1) In the trust management of traditional C2C, it is need to analyze features of reputation fraud, as well as the the existing problems of C2C reputation model, and finally to improve anti-fraud ability of trust management.(2) In the trust management of context-related C2C scene, transactions need to be dealt offline. Different context will have different perceptions of risk. The immature business model may prone to sparse feedback, and will lack of the feedback under the same context. It makes the trust assessment need to reduce the occasional trust, and to deal with different context reputation feedback.The main research work can be subdivided into the following four aspects. (1) Describe the traditional C2C trading scenario, and analyze the fraud in existing C2C reputation model. Then introduce the proposed C2CRep model in this work, including guidelines, parameter selection, the basis of various factors, and experimental evaluation.(2) Analyse the effect of buyers’review feedback to potential buyers in the C2C trust management, and proposes ranking method based on feedback reliability to show the reviews. In order to calculate the feedback reliability, the work has improved the weighted RFM customer value model to adapt to the C2C environment; introduced how to compute each index in RFM model; and then determined RFM weight by AHP method; finally used the customer value theory and the reputation to establish feedback reliability calculation model.(3) Describe the context-sensitive C2C trading scenario, and analyse risk assessment in different contexts. Drawing on information systems risk assessment methodology, the paper has proposed a method based on fuzzy set theory, including analysis of the context of risk hierarchy, quantitative risk analysis of all factors, comprehensive evaluation results of the context risk.(4) Because of context-sensitive C2C transactions prone to sparse feedback, and trust mapping in different context, the paper has proposed a way to filter false feedback rely on the feedback offset, calculated the stability by the amount of feedback. By combined with work context risk assessment, the work realized the trust mapping mechanism of different context through risk value, and constructed a risk perception of context in C2C trust management model.The main innovations of this work are as follow.(1) The paper proposed a C2CRep seller’s reputation model. The new model has added collusion factor to balance the number of transactions and concentrated feedback, added unsatisfied index to inhibit feedback blackmail, and improved price index to dynamically adapt to commodity’s personality. Compared with the traditional C2C model and classic SPORAS model, experimental results show that in low-price reputation deception, reputation collusion, and reputation slander, it can improve the ability to resist reputation fraud.(2) The paper proposed a feedback reliability compute model based on RFM. It has combined the theory of customer value, and improved RFM customer value model to fit C2C environment. Compared with the reputation based model, experimental results show that it can increase the cost of transaction time, frequency, money to obtain high ranking score through fraud trading.(3) The paper proposed a context risk related C2C trust model. It has constructed risk assessment of context in C2C model based on fuzzy set, and realized trust mapping through the risk indicator. It also has introduced a stable index in filtering inaccurate feedback by its offset to avoid accidental factor for insufficient amount of feedback. Simulation results show that, compared with the the traditional mean model and PeerTrust model, it improve the accuracy of trust evaluation, and can inhibit reputation fraud by low-risk context.The work has produced some fruit, which will enrich e-commerce trust management mechanism to some extent, and also further promote the C2C business applications development in future.更多还原