Closed santoshpalaskar77 closed 1 year ago
Hi @santoshpalaskar77 two ways you can do this currently:
make_evaluation_predictions
, by passing the training dataset, to get forecasts on the trailing prediction_length
window. Then you can do anything you want with forecasts, like visualizing them or measuring accuracy.backtest_metrics
, by passing the training dataset, to directly get evaluation metrics on the same trailing windows. Note that this relies on make_evaluation_predictions
, just directly does evaluations on top of it.Please be aware that the final window of each time series in the training data has probably been used for training purposes, therefore optimizing too much for accuracy there may lead to overfitting.
Description
(A clear and concise description of what the feature is.)