A semantic knowledge graph with insurance-specific entities can enable location underwriters to better assess their risk portfolio, by including all available information about a single geographic entity. However, the knowledge graph can, currently, only provide information that has either been ingested from the original data sources, or configured by an administrator. Our thesis states that there exist relationships between entities of the knowledge graph that are not represented by links, as yet, but can be predicted, either as unreported or potential (in the future) links.