Towards ontology-enhanced knowledge-based systems using ontology design patterns: application in credit risk management

Abstract

Recently, the amount of data generated on a daily basis has exponentially exploded. Tremendous efforts have been undertaken to provide the industry and research fields with expanding technologies for facilitating the collection, treatment, and sharing of massive data. This poses unprecedented challenges for the use of these technologies, especially in the credit risk management. The greatest challenge lies in collecting, retrieving, reusing and sharing this knowledge on one hand, and in creating new values for improving the competitiveness of financial institutions on the other hand. In the aftermath of the 2008 financial crisis, banking systems have shown a precarious fragility particularly in data architecture and information technology infrastructure. Moreover, inadequacy or non–existence of common vocabulary has created a gap in common semantics. Several solutions have been proposed, but the most efficient approach to respond to these issues is the use of ontologies. Ontologies are a natural fit for data knowledge representation, and data storing, retrieving, integrating, and reasoning. This work used the ontology for modeling the credit risk scorecard. We opted for the Krisnadhi & Hitzler methodology based on Ontology Design Patterns, which is a worked and tested example, to develop credit risk scorecard and applicant ontologies. The methodology supports collaborative construction and allows for reusability. We linked the credit risk scorecard to credit risk scorecard objectives and credit risk scoring decision support tool. Our credit risk scorecard ontology is then used to identify the corresponding tasks, subtasks, roles, actors, and resources of the affected business. Our ontology is validated using the methods and tools already tested. Firstly, we make the diagnosis and repair of the ontology for quality validation. Secondly, we reason the logical consistency of the ontology. Thirdly, we make queries answering for usability validation of the ontology. Our ontology is a shareable ontology that can be understood both by humans and computers alike. The major aim of this dissertation is to develop a rich, expandable, and re-usable credit risk scorecard based on ontology design patterns and in compliance with BCBS 239. The principal result of this work is laying the cornerstone for the credit risk management platform-based ontology, which remains open to the extent that it allows for constant accumulation of tacit knowledge and efficient polymerization of explicit knowledge.