Trust-based service selection and recommendation for online software marketplaces – TruSStReMark

Abstract

This dissertation proposes a framework (TruSStReMark - Trust-based Service Selection and Recommendation for Online Software Marketplaces) to model, quantify, and monitor trust of software services and to perform trust-based service selection and recommendations. It provides methods to analyze and aggregate external reviews, pertaining to specific QoS attributes, of software services by performing subjective logic-based operations. This framework, first, defines trust of a software service using theory of belief and extends the multi-level software specifications to represent the trust-based attributes. It, then, proposes enhancements to two prevalent algorithms for selecting and recommending software services from a marketplace. Finally, the performances of the enhanced selection and recommendation algorithms are improved by parallelizing them. When compared with the prevalent Content-based and Collaborative filtering-based approaches, the results show that, the TruSStReMark is able to produce better results in terms of quality measured using HR (Hit Ratio) and ARHR (Average Reciprocal Hit-Rank) metrics. In addition, the parallelized versions of the trust-based selection and recommendation algorithms improve the end-to-end runtime. The TruSStReMark will enable users to select services, which are trustworthy, from online software marketplaces and use them in composing quality-aware distributed systems.