ProblemsByVin's picture
Fix dataset card YAML + point viewer at CSV via configs
d4716a3 verified
|
Raw
History Blame Contribute Delete
1.87 kB
metadata
license: cc-by-4.0
language:
  - en
pretty_name: US Vehicle Failure-Mileage Distribution (NHTSA, 2005-2025)
tags:
  - automotive
  - vehicles
  - nhtsa
  - vehicle-reliability
  - recalls
  - failure-analysis
  - open-data
  - consumer-data
size_categories:
  - 1K<n<10K
configs:
  - config_name: default
    data_files:
      - split: train
        path: failure-mileage-distribution.csv

US Vehicle Failure-Mileage Distribution (NHTSA Owner Complaints, MY 2005–2025)

For every U.S. vehicle + component failure pattern with enough mileage-bearing complaints, the mileage at which owners report the failure — as quartiles (25th percentile / median / 75th percentile). ~9,100 (vehicle, component-category) clusters derived from the U.S. NHTSA owner-complaint record.

Mileage is parsed from NHTSA's ODI bulk flat-file, which carries odometer readings the public JSON API omits, validated to a 1,000–400,000-mile band. Quartiles are published only for clusters with ≥8 mileage-bearing complaints. No other public source aggregates complaint mileage at this resolution.

Columns

column meaning
year, make, model vehicle identity
component_category NHTSA component bucket (e.g. powertrain, engine, brakes)
complaint_count complaints in the cluster
mileage_q1 / mileage_median / mileage_q3 reported failure-mileage quartiles
mileage_sample_size complaints with a valid odometer reading

Source & methodology

Derived from public U.S. NHTSA owner complaints and recalls. NHTSA complaints are unverified consumer reports. Canonical version, full methodology, and the other datasets: https://problemsbyvin.com/data/ . Also archived with a DOI on Zenodo: https://doi.org/10.5281/zenodo.20648310

License

CC BY 4.0 — free to use with attribution to ProblemsByVin (https://problemsbyvin.com).