Studies Of Theory And Critical Approaches Of Users Perceived Trust Recommender System In B2C

With the development of E-commerce and shopping platform, online shoping isgradually accepted by people. Because of the goods are in great abundance, in orderto improve service quality the website needs a tool to provide goods for consumersquickly and precisely. Recommendation system is not only a way to improveconsumer’s decision quality and decision efficiency, but also help website increaseprofits and decrease costs. Therefore, recommender system has enormous value toresearch and apply.In research area there are enormous research achievements about recommendersystem. These findings not only improve the precise of recommendation but alsomeet consumer’s novel needs. However, recommender system should obtain thetrust from consumers as a precondition for realizing it’s role. If there is no trust, itwill make the effects of recommender on consumers and website big sell at adiscount. So, the purpose of this paper is to design a trust-based recommendersystem. Through interacting with system, the consumer will perceive benevolence,integrity and competence of system. The research process is based on theinformation system design theory. The research process and contents follows:We first constructed trust-based theory model through theoretical derivation.The model could explain consumer’s process of trust formation gradually throughinteracting with recommender system. The designed system reflects a supporting forconsumer’s purchase decision process. Consumer’s trust develops when one gainsknowledge about trustee on the process of interacting. In detail, we analized theeffects of consumer’s task-system function fit on willing to use initially based ontask-technology fit model. Howerver according to Elaboration Likelihood Model,active engagement is important for consumers to understand and evaluate system, soconsumers’s good experience is very important. Therefor we analized the effects ofcharacters of system, feedback, skill-challenge fit, on consumer’s good experiencesthrough flow experience theory. Furthermore we analized the effects of goodexperience on consumer’s perceived control and perceived enjoyment. Finally theconsumer’s perceived ease of use and perceived transparency affects consumer’strust about system. Based on the proposed theory model, we construct framework ofrecommender system and specified each sub-constructs. The functions of the systeminclude supporting consumer’s purchase decision process such as needs recognize,searching, evaluation. In addition to this, the system also give consumers rightfeedback at right time.In order to solve the problem of surppoting consumer’s need recognize, we takes advantage of components analysis and betweenness centrality and Q-measuresof whole network to find the topic preference of consumer and the critical “bridge”goods between two different topic preferences. Through these analysis, on onehands,we could predict the consumer’s needs precisely and diversly according tothe similarity and continuous characters of consumer’s need. On another hands, thewebsite will make consumers’s preference from current to other preference ingreately possible. Because of the network analysis will be done offline, the websitecould support consumer’s needs rcognize quickly, precisely and novel.In order to help consumers search and evaluate goods quickly and precisely, wepropose an approach of neighbor selection based on user classification. We firstdivide consumers into three types, which is expertise, trustworthy, similarity, andthen analized the important of them to target users, finally the proper neighborsmight be found for the target user. Through these neighbors of target user, thesystem could help consumers evaluate goods and get recommendation moreprecisely. Meanwhile, based on these method the website could set some feedbackinterface to promote consumers interact with system, finally improve consumer’sexperiences about system through trade-off explanation.Collaborative filtering system has been limited to some extent because of thesparsity of common rations between users. In the electronic commerce, it is difficultto find same purchase history between two consumer, therefor the sparsity problemis common. In order to improve the adaptiveness of system in sparsity envirenment,we first adopts Association rule mining to formalize similarity among competitivegoods based on considering consumption level, and then takes advantage ofcomponents analysis of whole network to find complementary similarity of thegoods for expanding common rating sets. the result shows that the sparsity problemhas been moderated to some extent and the accuracy and diversity of have beenimproved significantly.At last, it is important for recommender system to give consumers rightfeedback of outcome and explanation at right time automatically. In this case thesystem could get trust from consumers in great possiblely. Therefor we constructedthe architecture of personalized feedback agent and specified its critical componentsincluding objects, events, beliefs, plans. Finally designed the stratigic of plansselection.Overall, trust-based recommender system is designed following the researchline, that is, theoritical analysis, architecture design and functional analysis,algorithm realization and validation, the feedback design of outcome andexplanation. Many new ideas and methods are provided. On the one hand, theyenrich current researches of trust theory obout recommender system from trustformation pespective, on the other hand, it can be applied in developing practical trust-based recommender system.