Reducing search space for Web Service ranking using semantic logs and Semantic FP-Tree based association rule mining

Abstract

Ranking and Adaptation (used interchangeably) is often carried out using functional and non-functional information of Web Services. Such approaches are dependent on heavy and rich semantic descriptions as well as unstructured and scattered information about any past interactions between clients and Web Services. Existing approaches are either found to be focusing on semantic modeling and representation only, or using data mining and machine learning based approaches on unstructured and raw data to perform discovery and ranking. We propose a novel approach to allow semantically empowered representation of logs during Web Service execution and then use such logs to perform ranking and adaptation of discovered Web Services. We have found that combining both approaches together into a hybrid approach would enable formal representation of Web Services data which would boost data mining as well as machine learning based solutions to process such data. We have built Semantic FP-Trees based technique to perform association rule learning on functional and non-functional characteristics of Web Services. The process of automated execution of Web Services is improved in two steps, i.e., (1) we provide semantically formalized logs that maintain well-structured and formalized information about past interactions of Services Consumers and Web Services, (2) we perform an extended association rule mining on semantically formalized logs to find out any possible correlations that can used to pre-filter Web Services and reduce search space during the process of automated ranking and adaptation of Web Services. We have conducted comprehensive evaluation to demonstrate the efficiency, effectiveness and usability of our proposed approach.