A Knowledge Graph-based Architecture to Understand Complex Economic Systems : An Example of Canadian Banking and Real Estate Sectors

Modern development of technologies in Data and Analytics has made alternative architecture of complex economic systems possible. Increasingly commoditized models that were developed to accept just a handful of well-known variables as input parameters would need to adapt to a reality, where growth of data by far outpaces growth of economic models. We propose a Knowledge Graph (KG)-based architecture where vast number of data sources on multiple domains can be integrated into a single system, where sub-networks of Law, Economics, Public Health, and Business come together to help assess how scenario in one sub-system may evolve to substantial impact to another sub-system. Besides the conceptual architecture, we also highlight key engineering structures and blocking issues when a KG-based system is applied to address real world questions. In a case study, we use the Deep Learning approach on the KG to solve for a specific question of how interest rate influences the systemic health of Canadian Banking system. Our empirical results indicate over 30% improvement of the mean absolute percentage error (MAPE), meaningful evidence of how KG-based analysis can be both comprehensive and accurate