Time series forecasting of telecom revenue using Auto-ARIMA and walk-forward validation.
📈 Telecom Revenue Forecasting with Auto-ARIMA
Overview: This project focuses on forecasting daily telecom revenue using historical data. A Time Series forecasting approach was applied using an Auto-ARIMA model, and the model's performance was evaluated through walk-forward validation.
The goal was to build an accurate model that can predict future revenue patterns — a critical tool for business decision-making.
📂 Project Structure:
Exploratory Data Analysis (EDA)
Seasonal Decomposition
Stationarity Testing (ADF Test)
Auto-ARIMA modeling
Walk-forward validation
Performance evaluation
🛠 Technologies Used Python 3
Pandas — data manipulation
Matplotlib — visualization
pmdarima — Auto-ARIMA modeling
scikit-learn — metrics (MAE, RMSE, MAPE)
statsmodels — time series decomposition and ADF test
📊 Results Mean Absolute Error (MAE): ~0.60 million
Root Mean Squared Error (RMSE): ~0.75 million
Mean Absolute Percentage Error (MAPE): ~4.75%
The model consistently tracks the revenue patterns with high accuracy, making it a strong candidate for operational forecasting in telecom business environments.
✍️ Author Anso Michel Data Science & Analytics Enthusiast
📜 License: This project is licensed under the MIT License.