Abstract: The limitations of traditional service credible evaluation is studied in this paper.A novel method is proposed for service credible evaluation based on classification.To solve the three major problems for trust evaluation in service-oriented computing internet：data sparseness，cold start and scalability study.The novel credible evaluation method is proposed based on user and service characteristics of the similarity of classification and the trust is calculated by using the distance between the classes.The method of trust derivation is adopted by combing of traditional trust derivation and trust derivation based on classification.The novel method is proposed completely in this paper.Through theoretical analysis and experimental results show that：The new trust evaluation services can alleviate the user computing environment to a large extent，which including the data sparseness problem，new user participation and the existence of barriers of trust reasoning and evolution of trust exist rationality and scalability issues.Experimental results show that the method significantly improves the system’s trust evaluation accuracy.

Key words: Web services computing, trust level, clustering, evaluation, trust evolution

摘要： 围绕传统的服务可信评价存在的局限性展开研究，提出一种新的基于类的服务可信度评价方法。针对当前面向服务计算信任评价中普遍存在的３大问题：数据稀疏性、冷启动和可扩展性进行研究。文中的评价方法基于用户以及服务的相似性特性进行分类，以类间的距离作为信任度的差异进行计算，采用了类间的信任推理与传统信任推理相结合的信任推导方式。完整地给出了基本类的相关信任推理与评价的方法。通过理论分析与实验结果表明：新的信任评价方法在很大程度上可以有效缓解服务计算环境中存在用户信任数据稀疏的问题;新用户参与时存在的信任壁垒问题以及信任推理与演化方面存在的合理性与可扩展性问题。实验结果表明该方法显著地提高了系统信任评价的准确率。

关键词: 服务计算, 信任度, 聚类, 评价, 信任演化