How nkhanna94/road-accident-analysis-and-traffic-severity-prediction works
Project Type and Purpose
• This is a full-stack machine-learning application:
– Data ingestion, cleaning, EDA, and model training happen offline via Jupyter notebooks and scripts.
– A trained XGBoost model is served in a Flask web application for real-time severity predictions.
Main Architectural Components
a. Data Layer
• Raw data files (Data/*.csv)
• EDA & outlier‐removal notebooks (Data/*.ipynb)
b. Processing & Modeling Layer
• Model‐training notebooks (Models/*.ipynb) for DecisionTree, RandomForest, SVM, XGBoost, etc.
• Model artifacts (e.g., serialized XGBoost model – stored under WebApp/finaln.csv or a pickle file).
c. Deployment Layer (WebApp)
• Flask service (WebApp/app.py and root app.py)
• Templates (WebApp/templates/index.html, result.html)
• Static assets (WebApp/static/*.css)
d. Presentation Layer
• User’s web browser interacting with the Flask UI
Relationships & Data Flows
• Data Ingestion → Data Cleaning & EDA
(CSV files flow into notebooks/scripts under Data/)
• Cleaned data → Model Training
(Processed data fed into Models/*.ipynb)
• Model Training → Model Registry
(Best model (XGBoost) persisted into WebApp/finaln.csv or pickle)
• Model Registry → Flask API
(Flask app loads the model at startup)
• User Input → Flask API → Prediction Engine → Result
(Browser submits HTTP form to Flask; Flask passes features to the model; returns severity prediction; displayed in result.html)
Architectural Patterns & Design Principles
• Separation of concerns:
– Data processing and model training are decoupled from deployment code.
– Frontend templates & static assets isolated from Python logic.
• Reproducibility:
– Notebooks/documentation under Data/ and Models/ track experiments.
• Model selection pipeline:
– Multiple algorithms evaluated; best-performing XGBoost chosen for deployment.
Technologies & Frameworks to Annotate
• Python, Pandas, NumPy – data handling
• Matplotlib/Seaborn – visualizations (Graphs/)
• Scikit-learn, XGBoost – ML algorithms
• Flask – web service
• HTML/CSS – frontend