🛒 Retail Sales Forecasting & Inventory Optimization System

Industry-Level Retail Analytics Portfolio Project
Built with Python, Machine Learning, and Streamlit

---

📌 Project Overview

This project simulates a production-level retail analytics system used by companies like Walmart, Amazon, D-Mart, Flipkart, and Reliance Retail to:

• Forecast future sales using machine learning
• Optimize inventory levels to prevent stockouts and overstock
• Classify products using ABC analysis
• Generate automated reorder recommendations
• Visualize KPIs in an interactive Streamlit dashboard

The entire system is built on a synthetic but realistic retail dataset simulating 5 stores, 20 products, 5 categories, over 3 years of daily sales — including seasonality, promotions, and stockout events.

---

🔥 Business Problem Solved

Problem	Impact	Solution
Stockouts → Lost sales	₹ revenue loss	Safety Stock + ROP
Overstock → Tied capital	High holding cost	EOQ optimization
No demand visibility	Poor planning	ML Sales Forecast
No product prioritization	Wasteful resources	ABC Analysis
Manual inventory decisions	Human error	Automated Reorder Alerts

📊 Business KPIs (Simulated)

• Stockout Rate Reduction: ~60% after optimization
• Inventory Turnover Improvement: 2.1x → 3.8x
• Estimated Revenue Recovery: ₹12.5M from lost sales prevention
• Forecast Accuracy (MAPE): ~8–12%

---

🧠 Tech Stack

Component	Technology
Language	Python 3.9+
Data Processing	Pandas, NumPy
Machine Learning	Scikit-learn (Random Forest), XGBoost
Time Series	Statsmodels (ARIMA)
Visualization	Matplotlib, Seaborn, Plotly
Dashboard	Streamlit
Config	PyYAML
Model Persistence	Joblib

---

🏗️ Architecture

Raw Sales Data (Synthetic)
         │
         ▼
┌─────────────────────┐
│   Data Generation   │  ← Simulates stores, products, seasons, promos
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│   Preprocessing     │  ← Clean, validate, fill gaps, fix types
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│ Feature Engineering │  ← Lag, rolling, date, promotion features
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│  Forecasting Models │  ← Naive Baseline, Random Forest, XGBoost, ARIMA
│  + Model Comparison │  ← MAE, RMSE, R², MAPE
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│ Inventory           │  ← Safety Stock, ROP, EOQ, ABC, Reorder Alerts
│ Optimization        │
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│  Visualization +    │  ← 12 charts + Streamlit Dashboard
│  Dashboard          │
└─────────────────────┘

---

📁 Folder Structure

Retail-Sales-Forecasting-Inventory-Optimization/
│
├── app/
│   └── streamlit_app.py       ← Interactive dashboard (6 pages)
│
├── config/
│   └── config.yaml            ← All settings in one place
│
├── data/
│   ├── raw_sales_data.csv     ← Generated raw dataset
│   ├── processed_sales_data.csv ← Cleaned data
│   └── features_data.csv      ← Feature engineered data
│
├── docs/
│   └── project_explanation.md ← Detailed project documentation
│
├── images/                    ← All generated charts (12 PNG files)
│
├── models/
│   ├── rf_model.pkl           ← Saved Random Forest model
│   └── xgb_model.pkl          ← Saved XGBoost model
│
├── notebooks/
│   └── EDA_and_Walkthrough.ipynb ← Step-by-step notebook
│
├── outputs/
│   ├── forecast_results.csv   ← 30-day sales forecast
│   ├── test_predictions.csv   ← Model predictions on test set
│   ├── inventory_recommendations.csv ← Full inventory plan
│   ├── abc_analysis.csv       ← ABC product classification
│   ├── model_comparison.csv   ← Model performance table
│   └── feature_importance.csv ← Top features ranked
│
├── reports/                   ← Auto-generated reports
│
├── src/
│   ├── __init__.py
│   ├── utils.py               ← Shared utility functions
│   ├── data_generator.py      ← Synthetic data creation
│   ├── preprocessing.py       ← Data cleaning pipeline
│   ├── feature_engineering.py ← Feature creation pipeline
│   ├── forecasting.py         ← ML model training & forecast
│   ├── inventory_optimization.py ← Safety stock, EOQ, ROP, ABC
│   └── visualization.py       ← All charts and plots
│
├── main.py                    ← Master pipeline runner
├── requirements.txt           ← All dependencies
├── .gitignore
└── README.md

---

⚙️ Installation & Setup

Prerequisites

• Python 3.9 or higher
• pip package manager

Step 1: Clone the Repository

git clone https://github.com/YOUR_USERNAME/Retail-Sales-Forecasting-Inventory-Optimization.git
cd Retail-Sales-Forecasting-Inventory-Optimization

Step 2: Create Virtual Environment

Windows:

python -m venv venv
venv\Scripts\activate

Mac/Linux:

python3 -m venv venv
source venv/bin/activate

Step 3: Install Dependencies

pip install -r requirements.txt

Step 4: Run the Full Pipeline

python main.py

Step 5: Launch Dashboard

streamlit run app/streamlit_app.py

---

🗂️ Dataset Details

The dataset is synthetically generated to simulate a real retail environment.

Field	Description
date	Daily date (2022–2024)
store	5 store locations across India
category	Electronics, Grocery, Clothing, Home & Kitchen, Personal Care
product	20 products across all categories
unit_price	Product selling price (₹)
actual_sales	Units sold on that day
demand	Total demand (including lost sales)
revenue	Daily revenue (₹)
current_stock	Stock at end of day
is_promotion	Whether a discount was active
stockout_flag	Whether item was out of stock

Simulated Features:

• Seasonal spikes (Diwali, Christmas, Republic Day, Holi)
• Weekend demand boost
• Promotional discounts (5–35%)
• Stockout events and demand variability
• Supplier lead time (5–10 days)

---

🚀 How to Run

Run full pipeline (recommended):

python main.py

Run specific steps only:

python main.py --steps data preprocess features forecast inventory viz

Run individual modules:

python src/data_generator.py       # Generate data only
python src/preprocessing.py        # Clean data only
python src/feature_engineering.py  # Feature engineering only
python src/forecasting.py          # Train models only
python src/inventory_optimization.py # Inventory logic only
python src/visualization.py        # Generate all charts only

Launch dashboard:

streamlit run app/streamlit_app.py

---

📊 Results

Model Comparison

Model	MAE	RMSE	MAPE (%)	R²
Naive Baseline	~18.2	~24.5	~22.4	0.41
Random Forest	~5.8	~8.3	~9.2	0.89
XGBoost	~5.1	~7.6	~8.1	0.92

Inventory Optimization Summary

• Safety Stock calculation using Z-score (95% service level)
• Reorder Point = Avg Lead Time Demand + Safety Stock
• EOQ minimizes total ordering + holding costs
• ABC Analysis classifies products by revenue contribution

---

🖼️ Screenshots

Generated charts are saved in the images/ folder after running main.py

Chart	File
Dataset Overview	images/01_dataset_preview.png
Sales Trend	images/02_sales_trend.png
Category Sales	images/03_category_sales.png
Store Performance	images/04_store_performance.png
Seasonality Heatmap	images/05_seasonality_heatmap.png
Promotion Impact	images/06_promotion_impact.png
Forecast vs Actual	images/07_forecast_vs_actual.png
Future Forecast	images/08_future_forecast.png
ABC Analysis	images/09_abc_analysis.png
Inventory Dashboard	images/10_inventory_status.png
Feature Importance	images/11_feature_importance.png
Model Comparison	images/12_model_comparison.png

---

🔮 Future Improvements

• Real-time dashboard with live data API
• Multi-store regional demand forecasting
• Weather/event-based demand adjustment
• Price elasticity analysis
• FastAPI endpoint for prediction serving
• Automated email reorder alerts
• Anomaly detection for unusual sales patterns
• Integration with ERP/SAP systems

---

🎓 Learning Outcomes

After building this project, you will understand:

• End-to-end ML pipeline design for retail analytics
• Time-series feature engineering (lags, rolling, expanding)
• Model comparison and evaluation methodology
• Inventory optimization formulas (Safety Stock, ROP, EOQ)
• ABC Analysis for supply chain prioritization
• Streamlit dashboard development
• Professional GitHub project structure

---

👤 Author

Tufan Chowdhury
📧 chowdhuryrudra449@gmail.com
🔗 LinkedIn
🐙 GitHub

---

📄 License

This project is licensed under the MIT License — feel free to use it for learning and portfolio purposes.

---

⭐ If this project helped you, please star the repository!