Say I have a dataset with a monthly granularity with the following columns:
There is data for each month for all of 2016-2019, so I divide the data accordingly.
training_data: 2016-2017validation_data: 2018holdout_data: 2019If I have a fitted_model that is a ForecastingPipelineWrapper which is the best run from AutoML where I passed gave it training_data and validation_data.
Looking at the ForecastingPipelineWrapper class docstring documentation only serves to confuse me more. What is X_past, X_future, and Y_future?
How do I use the above dataframes with fitted_model.forecast() to manually validate model fit on the holdout_data dataframe?
The following notebook illustrates how to leverage y_past, x_past, y_future, x_future, and fitted_model.forecast in the bottom half, 'Forecasting away from training data'. https://github.com/Azure/MachineLearningNotebooks/blob/master/how-to-use-azureml/automated-machine-learning/forecasting-high-frequency/auto-ml-forecasting-function.ipynb
The notebook will be a much better guide to grasping these concepts than perhaps a docstring doc. Should you have any more questions or need clarity, let us know!