Weekly revenue and profit forecasting using SARIMA, Prophet, XGBoost and LSTM on Amazon e-commerce data.
This project builds an end-to-end time series forecasting system on real Amazon e-commerce order data. It combines four different forecasting models and provides an interactive Streamlit dashboard for analysis and prediction.
| Model | Description |
|---|---|
| SARIMA | Seasonal ARIMA for trend and seasonality |
| Prophet | Facebook Prophet with holiday effects |
| XGBoost | Gradient boosting with lag/rolling features |
| LSTM | Deep learning with long-term memory |
| Ensemble | Weighted combination of all models |
- Upload your own Amazon order CSV files or use built-in demo data
- Forecast revenue and/or profit for up to 26 weeks ahead
- Compare model performances (MAE, RMSE, MAPE)
- Confidence interval bands for uncertainty estimation
- Interactive Plotly charts
amazon-revenue-forecasting/
├── app.py # Main Streamlit application
├── requirements.txt # Python dependencies
├── runtime.txt # Python version (3.11)
└── README.md
git clone https://github.com/kaanaltay-hub/amazon-revenue-forecasting.git
cd amazon-revenue-forecasting
pip install -r requirements.txt
streamlit run app.py- Python 3.11
- streamlit, pandas, numpy, plotly
- scikit-learn, statsmodels
- prophet, xgboost, tensorflow
The app accepts two CSV files from Amazon Seller Central:
amazon_orders_2023_time_series.csvdf_time_series.csv
Or use the built-in Demo Data option from the sidebar.