Recommending Analytic Services for Population Health Studies Based on Feature Significance

Service-oriented thinking is one of the fastest growing paradigms in information technology, with relevance to many other disciplines. Service-oriented analytic workflows can bring together various analytic computing tools and compute resources offered as services to answer complex research questions. The current healthcare system in United States is experiencing fundamental transformation as it moves from a volume-based business to a value-based business. One strategy that healthcare organizations start to deploy is leveraging their healthcare data to gain insights for optimizing their operation. Therefore it is perfectly logical to extend the application of service-oriented analytic workflows to population health studies, as these rely on both medical expertise and processing of large data sets to serve end users of various backgrounds and skill sets. However, in the practical application of such service oriented approach, the user often finds it difficult to choose the right services or workflows that can help them to find the answers to their questions. To tackle this problem, we propose a heuristic recommendation method based on the feature significance. The user submits an enquiry, then based on which, the system will recommend the services and compositions that are likely to produce meaningful answers. In this paper, we will elaborate the interactions between different roles in a service oriented analytic system, develop the modeling to illustrate the relations among enquiry, features, services and workflows, propose the algorithm for service recommendation, architect the system and show a reference implementation of a prototype.