Forecasting with exogenous variables and ensemble learning in sktime
In this recipe, you’ll explore two additional techniques: incorporating exogenous variables and ensemble learning with multiple time-series forecasters. The EnsembleForecaster class in sktime allows you to combine several forecasters, where each forecaster is trained individually, each contributing to making the final prediction. The collective prediction is made by averaging the forecasts from each regressor, which is the default aggregation in EnsembleForecaster. Think of this as the power of the collective, or the “wisdom of the crowd,” to improve predictive accuracy. EnsembleForecaster supports other aggregations, including median, min, and max.In this recipe, you will build an ensemble using the same regressors from previous recipes, including linear regression, Random Forest, Gradient Boosting, KNN, and SVM. You will use the reduction technique explored in earlier recipes, with make_reduction...
