【摘要】 社会化商务中,消费者依赖在线口口相传建立感知信任,其本质是复杂网络上的多源信任融合问题。国内外学者对多源信任融合问题进行了大量研究,并以主观逻辑方法为代表形成了信任融合方法的研究体系。然而,由于社交网络中消费者感知信任的多源性和高度主观性以及用户生成内容的海量化,给多源信任融合模型带来量化难、实时处理难和融合难等问题。针对上述难题,提出"先聚类、后融合"的研究思路,先对海量推荐信息进行聚类,再融入感知信任的主观因素构建多源信任融合模型。首先,将推荐信息间的相似性作为节点关系,从社交网络中抽取出推荐信息相似性网络,用谱平分方法对聚簇进行划分,实现对推荐信息的聚类;其次,用网络属性度量感知信任的影响因素,从复杂网络视角出发,提出消费者感知信任定性因素的量化方法;再次,以多属性决策方法为基础,改进主观逻辑方法构建多源信任融合模型,从而将感知信任的影响因素融入主观逻辑方法,突破主观逻辑方法只考虑推荐信息和网络路径的局限性;最后,通过仿真实验对推荐信息实验数据进行聚类,并分析主观因素对感知信任意见空间的调节作用,验证模型的可行性。研究结果表明,研究模型能够快速划分推荐信息相似性网络,客观地量化感知信任的影响因素,使其融入信任度计算之中,且能够体现消费者感知信任的主观性和异质性。从仿真实验结果看,该模型能够有效解决大规模社会网络中推荐信息海量化问题,权威程度、从众行为和主体间亲密度等影响因素对信任度计算结果起调节作用。该模型将信任融合模型扩展到社会化商务领域,可以帮助商家评价已有消费群体对新消费者感知信任的影响力,为大规模网络中消费者感知信任的度量和预测提供新视角,为商家实时分析消费者感知信任意向和制定营销策略提供方法支持。对于难以建立新消费者信任的商家,可以通过制定激励机制来提高用户生成内容量,培养或引入权威人士作为明星节点,建立有主题的小社区等形式来促进主体间交流,增强消费者之间的亲密度,从而提高消费者感知信任。更多还原

【Abstract】 In social commerce,consumers establish perceived trust through online word-of-mouth( WOM),whose essence is multi-source trust fusion which has been fully analyzed by scholars worldwide and an analysis system has been formed based on subjective logic. However,due to multi-sources and high subjectivity of consumers’ perceived trust as well as massive user-generated content( UGC),there still exist some key issues in trust fusion model development in relation to quantitative processing,real-time processing and information fusion given the context of social commerce.We propose a method of "clustering before fusion"to establish the multi-source trust fusion model. First,this study takes similarity of recommendation as relationships between nodes to extract recommendation similarity network from social networks,and then partitions the recommendation similarity network with spectral bisection in order to cluster recommendation information.Second,we come up with the quantitative approach of perceived trust from the perspective of complex network. Third,we establish a multi-source trust fusion model by improving subjective logic based on multiple attribute decision making. The model integrates influential factors and breaks through the limits that subjective logic only takes into account the recommendations and network path. Finally,we verify the feasibility of the model. The data of recommendation experiments are clustered by simulation experiments and the mediating effects of subjective factors on perceived trust space are tested.This study indicates that the recommendation can be modular in similarity network rapidly with the use of our model mentioned above. In addition,measuring the influential factors objectively and taking account into trust degree calculation will deliver the subjectivity and heterogeneity of consumers. Based on the simulation results,massive recommendation information in social networks has been effectively addressed by our model. Meanwhile,the influential factors such as authority,herd behavior and closeness act a mediating effect in the calculation result of trust degree.Extending the use of trust fusion model into social commerce can evaluate the spreading effect from existing consumers to new consumers in perceived trust. Thus this model provides a new perspective for measuring and forecasting consumers’ perceived trust in massive social networks,and offers sellers a method in real-time analysis of consumers’ perceived trust intention and market strategy development. For the sellers with obstacles in new consumers’ trust establishing,raising the quantity of UGC will be a good choice. Developing stimulating mechanism,fostering or introducing authority as stars,building small communities to promote communication can be alternatives to finally enhance the closeness between consumers.更多还原