wwei0810/AirSafe_DB

✈️ Aircraft Accident Dataset (Normalized)

This dataset contains structured information about aircraft accidents, extracted and normalized from raw CSV data scraped from the Plane Crash Info website (http://www.planecrashinfo.com/) on December 27, 2024. The data has been processed by an AI model to ensure consistency and accuracy, and it is presented in a standardized JSON format.

Dataset Description

This dataset is derived from a collection of aircraft accident reports, originally in CSV format, obtained from the Plane Crash Info website. The raw data was scraped on December 27, 2024, with a total of 5,049 entries, and has been transformed into a structured JSON format with the help of an AI model (deepseek v3 by default). The normalization process ensures data quality and consistency, making it suitable for various analytical and research purposes.

Each accident record in the dataset includes the following information (where available):

• Incident Information: Date, time, location (city, country, airport, runway).
• Aircraft Information: Operator, flight number, route, aircraft type (manufacturer, model, variant), registration, construction number.
• Casualty Statistics: Total fatalities, breakdown of fatalities (crew, passengers, ground), and the original raw text describing the casualties.
• Accident Summary: A detailed description of the accident, the type of accident (e.g., ground collision), the phase of flight (e.g., landing), circumstances, and mission type.
• Collision Flag: Indicates if the accident involved multiple aircraft.

Data Structure

The dataset is structured as a collection of JSON objects, each representing a unique accident. Here is a simplified example of the data structure:

[
  {
    "incident_info": {
      "url": "https://www.planecrashinfo.com/2024/2024-1.htm",
      "date": "2024-01-02",
      "time": "17:45",
      "location": {
        "city": "Tokyo",
        "country": "Japan",
        "airport": "Tokyo Haneda Airport",
        "runway": "34R"
      }
    },
    "is_collision": true,
    "total_casualties": {
      "total": 5,
      "known_distribution": {
        "crew": 5,
        "passengers": 0,
        "ground": 0
      },
      "raw_text": "5 (passengers:0 crew:5)"
    },
    "aircraft": [
      {
        "operator": "Japan Airlines",
        "flight_number": "JL516",
        "route": {
          "from": "Sapporo",
          "to": "Tokyo"
        },
        "aircraft_info": {
          "raw_name": "Airbus A350-941",
          "manufacturer": "Airbus",
          "model": "A350",
          "variant": "941",
          "registration": "JA13XJ",
          "construction_number": "538"
        },
        "occupants": {
          "total": 379,
          "known_distribution": {
            "crew": 12,
            "passengers": 367
          }
        },
        "casualties": {
          "total": 0,
          "known_distribution": {
            "crew": 0,
            "passengers": 0
          },
          "raw_text": "No casualties in this aircraft"
        }
      },
      {
        "operator": "Japan Coast Guard",
        "flight_number": null,
        "route": {
          "from": "Tokyo",
          "to": "Nigata"
        },
        "aircraft_info": {
          "raw_name": "de Havilla DHC-8-315Q",
          "manufacturer": "de Havilland",
          "model": "DHC-8",
          "variant": "315Q",
          "registration": "JA722A",
          "construction_number": "656"
        },
        "occupants": {
          "total": 6,
          "known_distribution": {
            "crew": 6,
            "passengers": 0
          }
        },
        "casualties": {
          "total": 5,
          "known_distribution": {
            "crew": 5,
            "passengers": 0
          },
          "raw_text": "5 crew members died, captain survived"
        }
      }
    ],
    "accident_summary": {
      "description": "A Japan Airlines Airbus, collided with a Japan Coast Guard aircraft on the runway at Tokyo Haneda Airport. The Airbus was landing on runway 34R when it collided with the Coast Guard aircraft. Both aircraft burst into flames. The Airbus continued 5,500 ft. down the runway until it came to a stop on the right edge of the runway. Miraculously, all 367 passengers and 12 crew escaped and survived the crash on the Airbus. The captain of the Japan Coast Guard aircraft also survived, but 5 others did not. The DHC-8 was on its way to transport supplies to Niigata in response to the recent earthquake.",
      "type": "ground collision",
      "phase": "landing",
      "circumstances": [
        "runway collision",
        "fire"
      ],
      "mission_type": [
        "commercial flight",
        "coast guard mission"
      ]
    }
  }
]

Data Normalization and Quality

The data normalization process includes:

• Standardized Date and Time: Dates are formatted as YYYY-MM-DD and times as HH:MM (24-hour format).
• Structured Location Information: Location details are broken down into city, country, airport, and runway where available.
• Detailed Aircraft Information: Aircraft details include manufacturer, model, variant, registration, and construction number.
• Consistent Casualty Reporting: Casualty figures are categorized into crew, passengers, and ground fatalities, with a total count and the original raw text for reference.
• Comprehensive Accident Summary: Each accident is accompanied by a description, type, phase, circumstances, and mission type.
• JSON Schema Validation: Each record is validated against a predefined JSON schema to ensure data integrity.

Usage

This dataset is ideal for:

• Accident Analysis: Studying patterns, trends, and contributing factors in aircraft accidents.
• Safety Research: Investigating the effectiveness of safety measures and regulations.
• Machine Learning: Training models for accident prediction, risk assessment, and anomaly detection.
• Data Visualization: Creating informative visualizations and dashboards related to aviation safety.

Dataset Creation

This dataset was created using the ac_normalizer.py script. The script processes raw CSV data scraped from the Plane Crash Info website on December 27, 2024, and utilizes an AI model to extract and normalize the information.

ac_normalizer.py Script

The ac_normalizer.py script is a Python-based tool that leverages AI models to transform raw aircraft accident data from CSV files into a structured JSON format. It supports multi-threading for faster processing and allows resuming from checkpoints to handle large datasets efficiently.

Usage of the ac_normalizer.py Script:

python ac_normalizer.py -i <input_path> [options]

Required Arguments:

• -i, --input: Input CSV file path or directory containing CSV files.

Optional Arguments:

• -t, --threads: Number of threads to use (default: 4).
• -n, --no-logging: Disable logging output.
• -s, --save-interval: Save every N records (default: 10).
• -r, --resume: Resume from the last checkpoint.
• -o, --output-dir: Output directory path (default: clean_data_logs).
• -m, --model: Model to use for processing (default: deepseek/deepseek-chat).

Example Command:

python ac_normalizer.py -i data/accidents.csv -t 8 -s 20 -o output/

Acknowledgements

Special thanks to the Plane Crash Info website (http://www.planecrashinfo.com/) for providing the raw data used to create this dataset. We also acknowledge the developers of the AI models used for data normalization.