Places filtering
← All revisions
editorial(services): rewrite service-places-filtering (Track-B) — confirmed no Rust crate by this name exists, same phantom-service pattern as service-business-clustering; real implementation is the Python GIS pipeline in project-gis; genericized specific thresholds/buffer-distance/filename per public-repo-safety-review.md, matching service-business-clustering's already-established treatment; register-clean EN+ES
@@ -1,6 +1,6 @@ --- schema: foundry-doc-v1 title: "Places filtering service" title: "Places filtering" slug: service-places-filtering category: services type: topic @@ -9,48 +9,55 @@ quality: complete index_group: specialist-and-domain-services status: active audience: public short_description: "service-places filters raw civic and institutional data to retain only regional-grade facilities, so GIS tier rankings reflect institutional concentration." short_description: "A filtering step that keeps only regional-grade institutions from raw civic data, so GIS tier rankings reflect institutional concentration rather than every clinic and community facility." bcsc_class: public-disclosure-safe language_protocol: PROSE-TOPIC last_edited: 2026-08-01 last_edited: 2026-08-22 editor: pointsav-engineering paired_with: service-places-filtering.es.md cites: [] --- **Correction (2026-08-02, verified against canonical `origin/main`):** `service-places-filtering` (and its companion `service-business-clustering`) don't exist as real crates anywhere in the monorepo — already flagged twice this session (the GIS cluster merge, 2026-08-01; and `applications/app-orchestration-gis.md`, this same sweep). The real implementation is a Python pipeline in the separate `project-gis` archive, not a Rust service in this monorepo. **Flagged, not resolved.** [[retail-co-location-tier-methodology|GIS tier rankings]] depend on knowing where regional institutions sit, not where every local clinic or community facility sits. The platform's places-filtering step keeps only civic and institutional facilities that meet a regional scale — hospitals, universities, and validated major transport hubs above fixed size thresholds — and consolidates multi-point campus records into a single regional anchor. Local-service density is removed at this stage, so downstream rankings reflect institutional concentration rather than raw facility count. [[retail-co-location-tier-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. ## What the filter keeps ## Key Takeaways The filter applies fixed, structural thresholds rather than configurable parameters: a hospital must reach a minimum staffed-bed count, a university a minimum full-time-equivalent enrollment, and an airport must be a validated major regional hub rather than a general aviation facility. Institutions below these thresholds are dropped before any downstream scoring runs. - 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. ## Consolidating campus records ## Filtering Thresholds A large institutional campus often appears in raw open geospatial data as many separate points. The filter merges points that plausibly belong to the same physical campus into a single regional anchor with one unified centroid, preventing a single large institution from being counted many times over. The service applies attribute-weight filters to the raw civic data provided by [[service-fs-data-lake|`service-fs`]]: ## Where this fits in the pipeline - **Regional hospitals:** minimum capacity threshold (50+ staffed beds). - **Regional universities:** minimum enrollment threshold (1,000+ full-time equivalent students). - **Airports:** validated as major regional transport hubs; general aviation facilities are excluded. ## Spatial Aggregation Large institutional campuses frequently appear in raw open geospatial data as multiple separate points. `service-places` applies a 200 m spatial buffer to cluster these into a single regional anchor with a unified center of gravity, preventing over-counting of large campus footprints. ## Data Output The resulting `cleansed-places.jsonl` provides the regional anchor dataset that [[app-orchestration-gis]] uses when awarding final [[retail-co-location-tier-methodology|co-location tier rankings]]. Filtering runs as part of the same Python-based GIS pipeline documented in [[app-orchestration-gis]] — the code that turns raw geographic and business data into the regional co-location index — rather than as a separately deployed service. Its output feeds [[app-orchestration-gis]] alongside the retail clustering step from [[service-business-clustering]] when the pipeline assigns final co-location tiers. This article does not restate the pipeline's specific thresholds, buffer distances, or internal file names; the general pattern (drop sub-regional facilities, consolidate multi-point campuses to a single anchor) is the stable, public-facing part of the design. ## See also - [[service-fs-data-lake]] — GIS data lake supplying the raw civic and retail data - [[service-business-clustering]] — retail clustering service that consumes the filtered place data - [[app-orchestration-gis]] — orchestration layer that assembles tier rankings from both services - [[retail-co-location-tier-methodology]] — the tier methodology driving tier assignments - [[app-orchestration-gis]] — the pipeline this filtering step is part of - [[service-business-clustering]] — the retail clustering step that runs alongside it - [[service-fs-data-lake]] — the raw civic and retail data this step consumes - [[retail-co-location-tier-methodology]] — the tier methodology the filtered data feeds ## References