Automating component integration for Web-based data analysis

In this paper, we describe a component-based infrastructure for Web-based data analysis applications that reduces the need for custom integration programming. Our vision is that, to construct an application, scientists will give a specification of three kinds of components: the desired data source, a data analysis filter, and a visualization device. From this specification, our infrastructure will automatically generate the desired application. To support this vision, the infrastructure contains an intelligent component integration system (iCIS) that automates component retrieval, adaptation, and integration. Within iCIS, component knowledge is represented using formal mathematical specifications, to provide a common framework for reasoning about problem decomposition, component retrieval and adapter generation. Our current focus is on integration of data analysis filters with online data archives described using XML-based languages. We also discuss how iCIS is integrated with a 3-tier client/server application architecture to support in a heterogeneous and distributed computing environment.