TrafficSense: Smart Traffic Risk Prediction System

TrafficSense is an interactive decision-support dashboard designed to predict the severity of road accidents. By combining historical crash data, route parameters, and real-time weather conditions, it provides granular risk assessments for specific highway segments to aid in proactive traffic management.

Key Features

• Live Weather Integration: Fetches real-time temperature, wind, and precipitation data via the Open-Meteo API to dynamically adjust predictive risk models.
• Spatial Risk Mapping: Visualizes multi-segment highway corridors and historical accident hotspots using interactive Plotly maps.
• Explainable AI: Utilizes SHAP (SHapley Additive exPlanations) to provide human-readable insights into which road or weather factors most influenced a specific severity prediction.
• Data Persistence: Automatically logs new predictions to a CSV and allows users to append custom location segments directly from the user interface.

Tech Stack

• Frontend: Streamlit, Plotly
• Machine Learning: XGBoost, Scikit-learn, SHAP
• Data Processing: Pandas, NumPy
• External APIs: Open-Street Maps(For Map Display),Open-Meteo (Real-time Weather)

Project Structure

• app.py: The main Streamlit dashboard application.
• generate_coordinates.py: Script to interpolate and generate granular spatial mapping coordinates for road segments.
• notebooks/: Jupyter notebooks for exploratory data analysis (01), preprocessing (02), and model training (03).
• models/: Contains the serialized XGBoost model, label encoders, and evaluation metrics (.pkl and .json files).
• data/: Raw and processed dataset files, including prediction logs and location segments.

Installation and Setup

1. Clone the repository and navigate to the project directory:

   git clone https://github.com/Priyanshi-070207/Smart-Traffic-Risk-Prediction.git
   cd Smart-Traffic-Risk-Prediction

2. Install the required dependencies:

   pip install -r requirements.txt

3. Generate map coordinate segments (required for the dashboard map to render):

   python generate_coordinates.py

4. Launch the application:

   streamlit run app.py