Spatio-temporal knowledge graphs extend standard knowledge graphs by anchoring facts to both time and space, enabling queries about sequences, locations, and causality that static graphs cannot answer. The post explains the concept through three concrete use cases: patient journey modeling in healthcare (with a real-world reference to the German Center for Diabetes Research's 800M-node graph), sports play-by-play event analysis using football data, and fraud detection via temporal transaction chains. Additional domains covered include supply chain, epidemiology, maritime transport, and IoT sensor networks. The post also introduces context graphs as a related concept for modeling decision reasoning, and provides clear guidance on when spatio-temporal modeling is unnecessary — static catalogs, pure 'what' queries, or cases where location is irrelevant metadata.
Table of contents
Use case 2: Sports play-by-play event graphsGet Konrad Kaliciński’s stories in your inboxUse case 3: Fraud detection as temporal transaction chainsOther domains where this fitsA related concept: context graphsWhen you don’t need a spatio-temporal graphThe shiftSort: