Places filtering service
← All revisions
editorial(services+architecture+systems): enrich 3 thin articles + Phase 7 alias fix — Key Takeaways for service-business-clustering, service-places-filtering, mailbox-atomicity (EN+ES); aliases: [os-totebox] on totebox-os — Phase 6+7
@@ -19,6 +19,13 @@ cites: [] [[co-location-methodology|GIS tier rankings]] depend on knowing where regional institutions sit, not where every clinic and community college sits. **`service-places`** filters raw civic data to retain only regional-grade facilities — hospitals with at least 50 staffed beds, universities with at least 1,000 full-time-equivalent students, validated major regional transport hubs — and applies a 200 m spatial buffer to consolidate large institutional campuses into single regional anchors. Local-service density is filtered out at this stage; downstream rankings reflect institutional concentration rather than facility count. ## Key Takeaways - Regional-grade thresholds are hard-coded into the service: hospitals require ≥50 staffed beds, universities require ≥1,000 full-time-equivalent students, and general aviation facilities are excluded from airport scoring. These filters are structural, not configurable parameters. - A 200 m spatial buffer consolidates multi-point campus records — a large hospital campus often appears as dozens of OSM nodes — into a single regional anchor with a unified centroid. This prevents tier over-counting on large institutional footprints. - Output is `cleansed-places.jsonl`, consumed by `[[app-orchestration-gis]]` alongside the retail cluster dataset from `[[service-business-clustering]]` when assigning final co-location tiers. - The filtering stage is deliberately upstream of tier scoring. Once local-service density is removed here, all downstream scoring operates on a clean regional-institution signal. ## Filtering Thresholds The service applies attribute-weight filters to the raw civic data provided by [[service-fs-data-lake|`service-fs`]]: