Maritime Cargo Classification Pipeline | 2023-2025
Interactive charts, 3-year comparisons, top commodities, and key discoveries timeline.
Open Dashboard →Column schema evolution, processing order, rule execution matrix, and architecture diagrams.
Open Technical Flow →Complete system overview, architecture, file index, and next steps recommendations.
View Plan →Detailed Phase 10 results, commodity breakdown, and 106M tons added across 15 commodity types.
View Summary →Cross-year analysis, commodity trends, rule performance, and year-specific anomalies.
View Comparison →Harmonized cargo classification dictionary with HS codes, keywords, and 4-level taxonomy (Group → Commodity → Cargo → Cargo_Detail)
View on GitHub →Complete HS code hierarchy (HS2/HS4/HS6) with descriptions. Includes 7.6MB JSON with full classification structure.
Browse on GitHub →US port dictionary (68KB), ACE port codes, Schedule D, country codes, and waterway cross-references.
View on GitHub →Complete ship registry (5.4MB) with IMO numbers, vessel names, types, and carrier SCAC codes (43KB).
View on GitHub →Schedule B export/import codes (5.1MB total), SITC, NAICS, SIC, ATP, and end-use codes for trade classification.
Browse on GitHub →Complete source code, documentation, and all reference dictionaries available on GitHub.
View GitHub Repo →80.2% classified | 49.3% tonnage | 0.9% unclassified tonnage
⚠️ Large data files available via Google Drive
52.5% classified | 84.0% tonnage | LBK package rule: 275M tons!
⚠️ Large data files available via Google Drive
55.9% classified | 82.8% tonnage | Reefer spike: 5x increase
⚠️ Large data files available via Google Drive
501M tons captured from single package type indicator. Accounts for ~50% of all classified tonnage!
79M tons added in Phase 10 by targeting specific grades (BASRAH, KIRKUK, LIZA, TUPI)
User's "salt is just salt" intuition: 32.9M tons (13-19x better than complex variants)
Process ~37K new records/month with 2-3 minute runtime
Use 786K classified records as training set for automated rule refinement
Refine "Steel NOS" → "Cold Rolled Coils" for higher analytical value