Abstract: To improve the benign development of service ecosystem and to forecast potential perishing services individual,the service-tag bipartite graph was established by using complex network,and Service Similarity Network (SSN) based on service tag was built to quantize the competition relationship between services.The Feature Percentage Ranking (FPR) of services was extracted based on SSN properties.A statistic machine learning algorithm was introduced to distinguish perishing services from Web service ecosystem.The application of OpenAPI service ecosystem on ProgrammableWeb showed that the proposed method could predict the potential perishing service individual in system effectively and help users to select services for building up durable service compositions.

Key words: Web service ecosystem, perishing service individual, complex network, feature extraction, machine learning

摘要： 为了促进服务生态系统的良性发展,从服务生态系统的角度预测潜在的消亡服务个体：利用复杂网络的方法构建“服务—标签”二部图,建立基于服务标签的含权服务相似度网络,以量化服务个体间的竞争关系|在分析服务相似度网络中服务节点度的基础上,提出服务节点的特征百分比指标用于区分消亡和非消亡服务个体,进而给出一套统计机器学习的算法来预测Web服务生态系统中的消亡服务。在ProgrammableWeb上OpenAPI生态系统的实证分析表明,所提方法可以有效预测出服务生态系统中潜在消亡的服务个体,为服务使用者提供可靠的建议,从而保障服务组合的长期可用性。

关键词: Web服务生态系统, 消亡服务个体, 复杂网络, 特征提取, 机器学习