A Study Into The Key Technologies Of Reputation Reporting Architecture In Web Environment And Its Applications

The rapid development of modern service industry calls for a standardized reputation collecting and management service. In order to establish China's unofficial and commercial-oriented reputation architecture, the implementation of a Reputation Collecting and Management System is indispensable. The common problems in the traditional reputation architectures, e.g. the over-simplified design of evaluation grades, the poor anti-disturbing capability, all needs to be taken into consideration during the establishment of our reputation evaluation system.This dissertation will firstly introduce the reputation evaluation system and reputation aggregating algorithms adopted by our system We proposed to use the Service Quality Measurement Framework to settle the problem of over-simplified design of evaluation grades. Through dividing the evaluation process into evaluation rules, the evaluation result will be more concrete. To conquer the poor anti-disturbing capability problem in traditional reputation evaluation algorithms, we take measures to prevent the common swindle behaviors. For example, we use real-name mechanism to eliminate the negative effects of multi-identities and negative discrimination, make use of time and monetary factor to illustrate the variety of service qualities, and make advantage of some effective algorithms to reduce the effect of positive discrimination.Clustering Filtering Algorithm focus on the different distributing of fair evaluation results and positive discrimination results, and can reduce the reputation bias to some extent. However, when the fair results are low the algorithm's effect is dissatisfactory, and when the fair results are high, minus bias will appear. Therefore, we made more strict premises for the algorithm, introduced time factor and modified the distance function As a result, the effect of the algorithm improved dramatically.Social Network Based Reputation Algorithm considers the reputation calculation from a brand new point of view. It starts from studying the social relations of the trusting enterprise, and aggregates the recommendation opinions of these connected enterprises. What's more, based on the feedbacks, the algorithm can dynamically adjust the credibility of the enterprises' recommendations, thus improve the accuracy ofreputation service in the future. In this way, the algorithm can provide "studying" and"personalized" reputation service, so as to eliminate the positive discriminationgradually.In the last part of this dissertation we will introduce the design and implementation ofthe Reputation Collecting and Management System, and give an illustration of thesystem.