📘 E-Commerce Revenue Forecasting & Dynamic Pricing Optimization System

A complete ML-powered decision intelligence platform for e-commerce: revenue forecasting, elasticity-based price optimization, SHAP explainability, model drift detection, automated data cleaning, and Streamlit dashboard.

---

🧩 1. Project Overview

This project is an end-to-end decision intelligence platform for e-commerce businesses. It helps answer key business questions:

• How much revenue will we generate next week?
• Which categories are price-sensitive?
• How do discount strategies change revenue?
• How stable is our model—has drift occurred?
• Why is the model predicting what it predicts? (SHAP Explainability)

The system includes:

✔ Synthetic data generation ✔ Professional automated cleaning ✔ ML revenue forecasting ✔ Dynamic pricing engine ✔ Price elasticity estimation ✔ SHAP Explainability (beeswarm, force, bar, waterfall, comparison) ✔ Model drift detection ✔ Streamlit interactive dashboard ✔ Automated evaluation visualizations ✔ Full unit testing + Advanced test suite + CI/CD

---

🏢 2. Business Problem & Motivation

E-commerce companies need to make data-driven decisions on pricing, promotions, forecasting, and product-level strategy. This project simulates a real-company analytics workflow using machine learning + explainability.

---

🧬 3. Tech Stack

Languages & Core

Data & ML

Visualization

Models

App / Dashboard

Testing

Utilities

---

🧼 4. Data Cleaning Pipeline

Raw transactional data is cleaned automatically through app/cleaning.py, which handles:

• Missing values
• Outliers
• Wrong discounts
• Incorrect revenue
• Negative units
• Invalid promo flags
• Duplicate rows
• Date validation

✔ Ensures high-quality modeling data.

---

🖼️ 5. Screenshots

A visual walkthrough of the complete Streamlit dashboard.

1️⃣ Dashboard Overview

2️⃣ Data Quality Check

3️⃣ Model Evaluation

4️⃣ Historical Analytics

5️⃣ Forecast Simulator

6️⃣ Pricing Optimizer

7️⃣ SHAP Explainability

8️⃣ Model Drift Detection

---

📁 6. Project Folder Structure

Ecommerce-Revenue-Pricing-Optimizer/
│
├── app/
│   ├── cleaning.py               # Data cleaning pipeline
│   ├── data_loader.py            # Loads raw + processed data
│   ├── forecasting.py            # Model training, evaluation, plots
│   ├── pricing.py                # Elasticity + dynamic pricing
│   ├── insights.py               # Business insights & data quality
│   ├── drift_utils.py            # PSI-based drift detection
│   └── streamlit_app.py          # Full Streamlit dashboard
│
├── data/
│   ├── synthetic_generator.py
│   ├── raw
│   │   └── transactions.csv
│   └── processed
│       └── modeling_data.csv
│
├── models/
│   ├── revenue_model.pkl
│   └── elasticity.json
│
├── reports/
│   ├── csv
│   │   └── evaluation_report.txt
│   └── visuals
│       ├── actual_vs_predicted.png
│       ├── residual_distribution.png
│       ├── feature_importance.png
│       └── error_over_time.png
│
│
├── screenshots/
│   ├── dashboard_overview.png
│   ├── data_quality.png
│   ├── estimate_price_elasticity.png
│   ├── forecast_simulator.png
│   ├── historical_analytics.png
│   ├── model_drift.png
│   ├── model_evaluation.png
│   ├── pricing_optimizer.png
│   ├── shap_explainability.png
│   └──  train_model.png
│
├── tests/
│   ├── test_data_loader.py
│   ├── test_forecasting.py
│   ├── test_pricing.py
│   ├── test_app_model_file.py
│   ├── test_shap_explainability.py
│   ├── test_drift_detection.py
│   ├── test_visualizations.py
│   ├── test_forecast_stress.py
│   ├── test_pricing_stress.py
│   └── test_data_integrity.py
│
├── README.md
└── requirements.txt

---

📊 7. Streamlit Dashboard Overview

Your dashboard contains the following tabs:

1. 🧹 Data Quality Validation
2. 📈 Model Evaluation (Visuals)
3. 📊 Historical Analytics
4. 🔮 Forecast Simulator
5. 💰 Pricing Optimizer
6. 🧠 SHAP Explainability
7. ⚠️ Model Drift Detection

---

🚀 8. How to Run the Project

✅ STEP 1 — Clone the Repository

git clone https://github.com/girishshenoy16/Ecommerce-Revenue-Pricing-Optimizer.git
cd Ecommerce-Revenue-Pricing-Optimizer

✅ **STEP 2 — Create Virtual Environment

python -m venv venv
venv\Scripts\activate  # Windows

✅ **STEP 3 — Install Dependencies

python.exe -m pip install --upgrade pip
pip install -r requirements.txt

✅ STEP 4 — Generate Synthetic Data (once only)

python data/synthetic_generator.py

✅ STEP 5 — Clean the Data

python app/data_preprocessing.py

---

✅ STEP 6 — Launch the Streamlit App

streamlit run app/streamlit_app.py

---

⭐ STEP 7 — IMPORTANT FIRST RUN ACTIONS (DONE INSIDE STREAMLIT)

On first launch:

Open the Sidebar and click:

1. Train Revenue Model

Refer image below

2. Estimate Price Elasticity

Refer image below

This generates:

• revenue_model.pkl
• elasticity.json
• Evaluation plots (saved to /reports/)

After this, all tabs will work.

---

🧪 9. Testing the Project

Run all tests:

pytest -q

Run with verbose output:

pytest -v

Your test suite includes:

• Data loader tests
• Model training tests
• Pricing logic tests
• SHAP explainability tests
• Visualization tests
• Stress tests
• Drift detection tests

---

💡 10. Key Business Insights

• Electronics category has high price sensitivity
• Higher discounts → higher unit sales but lower margin
• Weekend & festive months (Oct–Nov) show spikes
• Promo share strongly influences daily revenue
• Model drift occurs during seasonal shifts (expected)

---

🚀 11. Future Enhancements

• SARIMA / Prophet forecasting
• Multi-product elasticity
• Automated retraining pipeline
• Inventory-aware price optimization
• Deployment on Streamlit Cloud