Scientific Workflow Recommendation Based on Service Knowledge Graph

Abstract

With the dramatically increasing emerging of external Web services, automatically creating workflows to satisfy the sophisticated requirements of users has become a significant issue. Most scientific workflow recommendation focus on mining association patterns between services in historical portfolios, including positive and negative rules, and recommending appropriate workflows based on derived patterns. However, due to the development of social network, several key social interactions are ignored which can enrich implicit associations of items and guide workflow recommendation. To tackle these problems, a Service Social Knowledge Graph (SSKG), including two types of entities service and developer and three types of relations isInk, isDlp and isFrd, is proposed to visually integrate and manage vital information which can facilitate workflows construction. Respectively, isInk shows the data flow between services, isDlp means the relation between a developer and his services and isFrd presents the friend relationships between developers. SSKG supplies indirect relations of services which inferred from isDlp and isFrnd. From SSKG, we extract several positive and negative rules to estimate the feasibility of composing services that the positive rules promote service composition and negative rules hinder the cooperation of services. According to the overall effects of rules, the $A^{*}$ and the Yen’s method are used to recommend workflows to users. We have conducted extensive experiments with real-world data. Results indicate that the accuracy and efficiency of our proposed method outperform the classical and state-of-the-art methods.