🛒 Walmart Sales Prediction Project

This project implements an end-to-end machine learning pipeline to predict weekly sales. It includes data preprocessing, feature engineering, model selection, and optimization using feature importance.

---

📁 Project Structure

• datacleaning.ipynb → Data ingestion, merging, preprocessing
• modeltraining.ipynb → Model training, evaluation, optimization
• models/
  • best_model_top.pkl
  • encoder_top.pkl
  • model_top_features.pkl
  • raw_top_features.pkl

---

🛠️ Data Cleaning & Preprocessing

The dataset is built by combining:

• features.csv
• stores.csv
• train.csv

Key Steps

• Dataset Merging

  • Joined on Store, Date, IsHoliday (left join)

• Missing Value Handling

  • Filled MarkDown columns with 0
  • Applied forward fill (ffill) for:
    • CPI
    • Unemployment (grouped by Store)

• Feature Engineering

  • Extracted:
    • Year
    • Month
    • Week (from Date column)

• Data Transformation

  • Converted IsHoliday to integer (0/1)

---

🤖 Model Training & Evaluation

Data Splitting

• Chronological split (80/20)
• Ensures validation on future data (real-world scenario)

---

📊 Model Performance

Model	MAE	RMSE	R²
Random Forest	1813.39	3764.04	0.9706
XGBoost	3459.77	5676.10	0.9332
Linear Regression	14542.03	20898.35	0.0950

Insight:
Random Forest performs the best, capturing ~97% of the variance.

---

⚙️ Feature Importance & Optimization

Top important features identified using permutation importance:

• Dept
• Size
• Store
• CPI
• Week
• Unemployment
• Type

The model was retrained using only these features.

Final Performance:

• R² Score: 0.9711

---

🚀 Deployment Artifacts

Saved in models/ for production use:

• best_model_top.pkl → Trained model
• encoder_top.pkl → One-hot encoder
• model_top_features.pkl → Processed features
• raw_top_features.pkl → Input features

---

🧠 Tech Stack

• Python
• Pandas, NumPy
• Scikit-learn
• XGBoost
• Jupyter Notebook

---

📌 Future Improvements

• Hyperparameter tuning (GridSearch, Optuna)
• Time-series models (ARIMA, Prophet)
• Streamlit / FastAPI deployment
• Real-time prediction pipeline

---

⭐ Summary

A complete ML pipeline for sales forecasting with strong performance using Random Forest and feature optimization.