Abstract
The Data Warehouse (DW) design remains a great challenge process for DW designers. As well, so far, there is no strong method to support the requirements analysis process in DW projects. The literature approaches try to solve this tedious and important issue; however, many of these approaches ignore or bypass the requirements elicitation phase. In this paper, we propose a method to generate multidimensional schemas from decisional requirements. We elected natural language (NL) like syntax for expressing decisional/business users' needs. Our approach distinguishes from existing ones in that it: i) is NLbased for requirements elicitation; ii) uses a matrix representation to normalize users' requirements, iii) automates the generation of star schemas relying on eight specific heuristics. We developed SSReq (Star Schemas from Requirements) prototype to demonstrate the feasibility of our approach illustrated with a real case study.