Traffic Flow and Road Safety Analysis Project

This repository now includes an actual runnable implementation (not only documentation) for traffic accident analysis with:

• data generation/loading,
• data cleaning + feature engineering,
• accident frequency modelling,
• accident severity modelling,
• metrics export for reporting.

Project Files

• Traffic_Flow_Road_Safety_Project_Guide.md — conceptual step-by-step project guide.
• src/generate_sample_data.py — creates a synthetic accident dataset if you do not have one yet.
• src/data_prep.py — cleans data and creates processed datasets.
• src/modeling_frequency.py — fits Poisson + Negative Binomial count models.
• src/modeling_severity.py — fits Logistic Regression severity model.
• src/run_pipeline.py — end-to-end runner.

Expected Input Data

Place your real dataset at:

data/raw/accidents.csv

Minimum useful columns:

• accident_id, date_time, segment_id
• road_type, intersection_type, lanes, speed_limit
• traffic_volume, weather, visibility, lighting
• vehicle_type, driver_age_group, injury_severity, injuries, fatalities

If the file is missing, run_pipeline.py automatically generates sample data.

Quick Start

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python src/run_pipeline.py

Outputs

After running, you will get:

• data/processed/accidents_clean.csv
• data/processed/frequency_dataset.csv
• outputs/model_metrics/frequency_metrics.json
• outputs/model_metrics/severity_metrics.json

These outputs can be directly used in your report’s Results and Discussion sections.