Government Record Linkage 2D — Fellegi-Sunter Score Explorer

Government data analytics rarely starts with a clean, unified dataset — it starts with dozens of agency databases that all describe the same citizens under slightly different names, addresses and typos, with no common key to join on. This simulator implements the actual algorithm used to solve that: Fellegi–Sunter probabilistic record linkage. Synthetic citizen records are generated for two agencies, corrupted by a tunable data-entry error rate, and every Agency-A-to-Agency-B pair is scored by comparing name, date of birth, address and a 4-digit ID suffix, each weighted by its real discriminating power. Drag the decision threshold and watch precision and recall trade off live — in the pannable/zoomable link diagram above and in the actual score-distribution histogram below it — as declared matches, correct in green and false in red, connect the two record columns.