Upon completion of this certificate, learners will possess the practical skills to implement XGBoost for time series forecasting tasks.
- Explain the core mechanics and advantages of the XGBoost algorithm compared to traditional decision tree methods.
- Identify key characteristics of time series data, including trends, seasonality, and stationarity.
- Prepare raw time series datasets by handling missing values and formatting features for model input.
- Construct lagged features and rolling statistics to transform time series data into a supervised learning problem.
- Train a basic gradient boosting model using the XGBoost library to predict future time series values.
- Evaluate model performance using standard regression metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE).
- Tune basic hyperparameters to improve the accuracy of the forecasting model.