📈 Sales Forecasting — Walmart Store Sales Prediction

🛒 A machine learning project that predicts weekly sales for Walmart stores using historical sales data, store features, and external factors like holidays, temperature, and fuel prices — helping retail businesses optimize inventory and planning.

---

🎯 Problem Statement

Walmart operates 45 stores across different regions. Accurate sales forecasting helps:

• ❌ Avoid overstock — wasted inventory
• ❌ Avoid understock — lost sales
• ✅ Optimize supply chain decisions
• ✅ Plan staffing and promotions

Goal: Predict weekly department-level sales for each store.

---

📂 Dataset — Walmart Store Sales (Kaggle)

File	Description	Size
train.csv	Historical weekly sales per store/dept	Training data
test.csv	Stores/depts to predict	Test data
features.csv	External features per store per week	Store metadata
stores.csv	Store type and size	45 stores

Key Features

Feature	Description
Store	Store number (1-45)
Dept	Department number
Date	Week of sales
Weekly_Sales	Target — sales in USD
IsHoliday	Holiday week flag
Temperature	Regional temperature
Fuel_Price	Fuel cost in the region
CPI	Consumer Price Index
Unemployment	Unemployment rate
MarkDown1-5	Promotional markdown data
Type	Store type (A/B/C)
Size	Store size in sq ft

---

🔄 Pipeline

Data Loading (train + features + stores CSV)
        ↓
Exploratory Data Analysis
├── Sales trends over time
├── Holiday vs non-holiday impact
└── Store type & size analysis
        ↓
Feature Engineering
├── Date features (week, month, year)
├── Holiday flag encoding
├── Store type encoding
└── Lag features for time series
        ↓
Model Training
├── Linear Regression (baseline)
├── Random Forest Regressor
└── Gradient Boosting / XGBoost
        ↓
Evaluation
├── MAE — Mean Absolute Error
├── RMSE — Root Mean Squared Error
└── WMAE — Weighted MAE (holidays weighted 5x)

---

🧠 Key Insights from Data

• 🎄 Holiday weeks have significantly higher sales
• 🏪 Type A stores (largest) generate highest revenue
• 📅 End of year (Nov-Dec) shows consistent sales spikes
• 📉 Markdowns correlate with increased sales activity
• 🌡️ Temperature & fuel price have minor but measurable impact

---

🛠️ Tech Stack

Layer	Technology
Language	Python 3.x
Data Processing	Pandas, NumPy
Visualization	Matplotlib, Seaborn
ML Models	Scikit-learn
Notebook	Jupyter Notebook

---

🚀 Getting Started

# Clone the repo
git clone https://github.com/tashfeen786/SalesForecasting.git
cd SalesForecasting

# Install dependencies
pip install pandas numpy matplotlib seaborn scikit-learn jupyter

# Run the notebook
jupyter notebook Task_07_SalesForecast.ipynb

---

🏗️ Project Structure

SalesForecasting/
│
├── Task_07_SalesForecast.ipynb    # Main analysis & modeling notebook
├── Task_07_SalesForecasting.pdf   # PDF export
├── train.csv                      # Training data
├── test.csv                       # Test data
├── features.csv                   # External store features
├── stores.csv                     # Store metadata
└── README.md

---

🔮 Future Improvements

• ARIMA / SARIMA — time series specific models
• XGBoost / LightGBM — better accuracy
• Feature importance analysis with SHAP
• Streamlit dashboard for interactive forecasting
• Cross-validation with time-series split

---

👨‍💻 Author

Tashfeen Aziz — AI/ML Engineer & Python Developer

---

⭐ If you found this project helpful, please give it a star!