Data Vault Mappings to Dimensional Model Using Schema Matching

Résumé En

In data warehousing, business driven development defines data requirements to fulfill reporting needs. A data warehouse stores current and historical data in one single place. Data warehouse architecture consists of several layers and each has its own purpose. A staging layer is a data storage area to assists data loadings, a data vault modelled layer is the persistent storage that integrates data and stores the history, whereas publish layer presents data using a vocabulary that is familiar to the information users. By following the process which is driven by business requirements and starts with publish layer structure, this creates a situation where manual work requires a specialist, who knows the data vault model. Our goal is to reduce the number of entities that can be selected in a transformation so that the individual developer does not need to know the whole solution, but can focus on a subset of entities (partial schema). In this paper, we present two different schema matchers, one based on attribute names, and another based on data flow mapping information. Schema matching based on data flow mappings is a novel addition to current schema matching literature. Through the example of Northwind, we show how these two different matchers affect the formation of a partial schema for transformation source entities. Based on our experiment with Northwind we conclude that combining schema matching algorithms produces correct entities in the partial schema.