一种面向 Mashup 应用的 API 推荐方法-《计算机技术与发展》

文章信息/Info

Author(s):: LI Xin; School of Computer and Information Engineering,Henan University,Kaifeng 475000,China

Keywords:: invalid service clustering; service recommendation; latent Dirichlet allocation; Word2Vec; Jaccard similarity coefficient

摘要:: 在动态变化的网络环境下,来自组合服务或 Mashup 的服务可能不可用或失效。此外,随着越来越多的服务在互联网上,用户很难找到他们想要的服务。服务聚类是辅助服务发现的一种重要方法,但是已有的服务发现方法没有考虑服务本身的规格(Specs)信息。针对上述问题,提出了一种面向 Mashup 应用的 API 推荐方法,通过 LDA 主题聚类对 API描述信息进行聚类,通过 Word2Vec 以及相似度计算从功能相似性的角度判断失效 API 所属主题类簇,在该主题类簇下通过失效 API 的规格信息进行进一步筛选, 将失效 API 与筛选后的主题类簇中的 API 进行相似度判断,从而得到推荐 API 的集合。最后以 ProgrammableWeb 网站上的真实数据进行了实验,与传统的 K-means、TF-IDF 方法相比,实验结果表明该方法在一定程度上可以提高推荐结果的准确性。

Abstract:: In the dynamic network environment,services from composite services or Mashup may not be available or fail. In addition,with more and more services on the Internet,it is difficult for users to find the services they want. Service clustering is an important method to assist service discovery,but the existing service discovery methods do not take into account the specs information of the service itself. Therefore, we propose an API recommendation method for Mashup application. We cluster the API description information through LDA topic clustering, and judge the subject cluster of the invalid API from the perspective of functional similarity through Word2Vec and similarity calculation. Under the subject cluster,we further screen the specification information of the fault API, and compare the invalid API with the API in the selected topic cluster. Finally, the real data on the programmable web site is tested.Compared with the traditional K-means and TF-IDF methods,the experiment shows that the proposed method can improve the accuracy of the recommended results to a certain extent.