AN AUTOMATED APPROACH TO SCHEMA MAPPING GENERATION FOR DATA WAREHOUSING

Abstract

In data warehousing, the source schemas are defined independently from the warehouse schemas, which are typically designed based on the information need of the warehouse users. The mappings between the source and warehouse schemas are also determined manually. Typically, more than one mapping between the warehouse schema and the source schemas is possible and the designer might miss the most appropriate mapping from the viewpoint of updates and maintenance of the warehouse. This dissertation focuses on the automated generation of the mappings between the source and the warehouse schemas that generates a complete list of mappings from which the warehouse designer can choose the appropriate mapping. This work discovers mappings between the schema elements based on linguistic and constrained-based matching of.element-level of schema-only based matching approach in the relational domain. Further, this dissertation designs a solution for automating the process of schema matching and mapping generation for the warehouse schemas pertaining to data warehouses for the relational domain handling SELECT-PROJECT-JOIN, UNION and INTERSECTION mappings. The design of the algorithm and the implementation issues, including the handling of homonyms, synonyms, derived attributes and the attribute mappings have been presented. Also the performance issues and ways to improve the performance of the algorithm including parallelization have been discussed thoroughly. Finally, the system developed in