DavisM1212/NHTSA-ODI-Complaint-Analytics
This project uses the NHTSA ODI consumer complaints to develop a method for identifying increasing safety concerns, prioritizing urgent complaints for review, and turning identified patterns into clear recommendations for investigation. The focus is on delivering real, practical value by combining multiple approaches.
What's novel
This project uses the NHTSA ODI consumer complaints to develop a method for identifying increasing safety concerns, prioritizing urgent complaints for review, and turning identified patterns into clear recommendations for investigation. The focus is on delivering real, practical value by combining multiple approaches.
Code Analysis
8 files read · 3 roundsEnd-to-end NHTSA ODI complaint analytics pipeline that ingests raw complaint data, cleans it into task-specific tables, and runs three locked ML systems (severity urgency scoring, component routing, and NLP early-warning topic monitoring) to produce reproducible benchmark artifacts, watchlists, and
Strengths
Genuinely deep implementation with multiple model families (CatBoost, SGD, Logistic, NMF), proper time-aware splits with locked holdout discipline, domain-specific text preprocessing with extensive regex patterns for NHTSA complaint artifacts, and sophisticated calibration (isotonic, power, sigmoid) with review-budget operational analysis. The architecture is clean with clear separation between config contracts, data I/O, preprocessing, modeling, and reporting layers.
Weaknesses
Test coverage is thin relative to the codebase size (9 test files for 88 source files), and several core modules like the NLP early-warning system (73KB) and cleaning pipeline (56KB) have limited test depth. The codebase has significant duplication of text preprocessing patterns across modules, and the 'locked' constants approach, while intentional, makes the system harder to adapt without touching core files.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
9
Contributors
1
Files
49
Active weeks
2
Repository
Language
Python
Stars
1
Forks
0
License
—