Deriving Data Modeling Requirements Using Topic Modelling in Cancer Care Context: The Case of the iCanViz Project

Abstract

The Development of a data model for the iCanViz cancer navigator, a smart cancer incidence analytics project, proved difficult due to the large body of resources to be analyzed. In this paper, we propose an approach to design, develop, and evaluate the data models for projects with large bodies of resources. The proposed approach utilizes topic modeling for data requirement discovery to analyze the semantic structures of thousands of documents and published articles and derive a reference conceptual data model for data strategy use-cases (DS-UCs) for cancer survival and deaths. Using this approach, we are able to process and analyze a great number of resources and map their automatically discovered topics into data model entities. Using the knowledge of a field expert, the derived model was evaluated and adjusted to arrive at a final model that captures the data requirements of the studied DS-UCs. The developed approach proved to be an effective benchmark for data modeling of a complex domain like cancer care, for which the validation of a domain expert is shown to be a critical element.