Skip to content
Open sourceAI / MLMachine learning

Traffic Accident Severity Prediction

Route-risk prediction trained on 100,000+ accident records

Role
Data science & backend
Period
2025
Status
Open source
Links
GitHub ↗

By the numbers

100K+
accident records
85%+
accuracy
3
models compared

Overview

Route-risk prediction trained on 100,000+ historical accident records, issuing real-time alerts at 85%+ accuracy using OpenRouteService and weather APIs.

  • Logistic regression, Random Forest and XGBoost compared; evaluated on accuracy, F1 and ROC-AUC.
  • A Flask API joins route segments with live weather and returns a risk score; a React front end shows it on the map.

01

Problem

Most factors that determine accident severity (weather, hour, road type, visibility) are known before the trip starts. The question: can that information produce a route-based, preventive alert for the driver or the authority?

02

Approach

More than 100,000 historical accident records were cleaned; new features were derived from datetime, weather and location fields. Logistic regression, Random Forest and XGBoost were trained and compared on accuracy, precision, recall, F1 and ROC-AUC.

A Flask REST API splits the OpenRouteService route into segments and predicts for each one with current weather. A React front end colours the risky segments on the map.

03

Architecture

  1. Data
    • 100K+ kaza kaydı
    • Öznitelik mühendisliği
  2. Model
    • Scikit-Learn
    • XGBoost
    • Değerlendirme
  3. Serving
    • Flask API
    • OpenRouteService
    • Hava durumu API
  4. UI
    • React harita
↓ Data flows top to bottom

04

Key decisions

  1. 01

    Split the route into segments

    A single 'this route is risky' score is useless. Segment-level prediction shows which part is risky and why.

Outcome

The best model reached over 85% accuracy. Experiments are reproducible in Jupyter notebooks; the API and the front end live in separate folders in the repository.

Related projects

  • AI / ML
    Open source

    Appliances Energy Prediction

    Comparing regression models for smart-home energy consumption

    Predicting appliance energy use from 19,735 ten-minute readings in the UCI dataset. Random Forest, XGBoost, LightGBM and linear regression were compared; LightGBM led with R² ≈ 0.76.

    19.735
    sensor rows
    0.76
    R² (LightGBM)
    • Python
    • Pandas
    • Scikit-Learn
    • LightGBM
    • XGBoost
    • Jupyter
    Open
    GitHub ↗
  • AI / ML
    Open source

    Mobile Price Classification

    Classifying phone price tiers from hardware specifications

    A classification study predicting a phone's price tier from hardware specs such as RAM, battery, screen and camera; includes a data-collection script, Jupyter analysis and a containerized prediction service.

    • Python
    • Pandas
    • Scikit-Learn
    • Jupyter
    • Docker
    Open
    GitHub ↗
  • AI / ML
    Live

    ChefsStack

    Recipe platform with a vector-based recommendation engine

    A recipe and recommendation platform where suggestions come from vector similarity instead of hand-written rules, and user interactions flow through Kafka and Redis into personalized feeds.

    <50ms
    similarity query
    1.000+
    interaction events processed
    • Python
    • FastAPI
    • PostgreSQL
    • pgvector
    • Redis
    • Kafka
    Open
    Visit site ↗