engleski

Ontology-supported schema enrichment of a relational data warehouse with multidimensional concepts from document-oriented data source

Acquisition of external data sources enables business enterprises to improve their predictive analytics and decision-making by increasing the pool of data on which they predicate their predictive models. Their existing analytical data are usually stored in a relational data warehouse, i.e., a centralized data repository. The popularity of document-oriented databases envisages the availability of many datasets in JSON format, but integrating such data into a relational data warehouse environment is problematic. Document-oriented data are semi-structured, lack an explicit schema definition, and are often permeated by heterogeneities making them schematically inconsistent. There is no consensus on the schema and reconciling it with a relational data warehouse schema poses problems. In this thesis, Semantic Web ontologies are used as a reconciliation medium, meaning that ontologies represent the relational data warehouse schema and a document-oriented dataset schema. This thesis provides a procedure for ontology-supported semi-automated enrichment of a data warehouse star schema with multidimensional concepts derived from document-oriented data. Another contribution of this thesis is a method for extracting an ontology from document-oriented data. This extracted ontology contains relationship cardinality information, crucial for multidimensional modeling. The procedure and method proposed in the thesis were verified using synthetic and real datasets. The evaluation results show that the ontology extraction method yields an ontology more suitable for multidimensional modeling than the existing approaches. The enrichment procedure successfully produces an ontological representation of a relational data warehouse’s enriched star schema.

ontology ; ontology-supported data integration ; relational data warehouse ; star schema ; star schema enrichment ; document-oriented data ; JSON ; ontology extraction

nije evidentirano