🚀 End-to-End ML Pipeline for Dynamic Price Optimization

📖 Overview

A production-ready Dynamic Pricing System that recommends optimal product prices using market demand, competitor pricing, and time trends. It demonstrates end-to-end ML engineering: data ingestion, feature engineering, model training, deployment, and interactive demo.

🏗️ Architecture

1. Data Ingestion – Load sales & competitor data.
2. Feature Engineering – Create rolling averages & store in PostgreSQL.
3. Model Training – Train & track experiments with MLflow.
4. Model Serving – Deploy best model with FastAPI.
5. Containerization – Dockerize for consistency.

🛠️ Tech Stack

• Python 3.9+, Pandas, NumPy, Scikit-learn, XGBoost
• Experiment Tracking: MLflow
• Database: PostgreSQL / SQLite
• API Framework: FastAPI
• Containerization: Docker

📁 Repo Structure

dynamic-pricing-ml-pipeline/
├── data/ (raw, processed)
├── notebooks/
├── src/ (data_processing, modeling, api, demo)
├── models/
├── docker/
├── requirements.txt
└── README.md

▶️ Setup & Run

# Clone repo & install dependencies
git clone https://github.com/your-username/dynamic-pricing-ml-pipeline.git
cd dynamic-pricing-ml-pipeline
pip install -r requirements.txt

# Run data pipeline
python src/data_processing/clean_data.py
python src/data_processing/feature_engineering.py

# Train model
python src/modeling/train_model.py
mlflow ui  # view results

# Run API with Docker
docker build -t price-api -f docker/Dockerfile .
docker run -p 8000:80 price-api

# Streamlit demo
streamlit run src/demo/app.py

📊 Results

• XGBoost Model Performance

  • MAE: ~$1.20
  • RMSE: ~$1.85
  • R²: 0.89

🔮 Future Work

• Real-time data ingestion (Kafka/AWS Kinesis)
• Orchestration with Airflow/Prefect
• Model monitoring (Evidently AI / WhyLabs)
• Additional features (inventory, weather, campaigns)

---

👨‍💻 Developer: [Mohd Fardeen Khan] 🔗 GitHub: [Fardeen-785] |