When calling gluonts.evaluation.Evaluator(), include all the metrics that are calculated in the aggregate metrics in the metrics per time series, e.g. RMSE #3152
Some of the metrics are included in the aggregated metrics, but not the per time series metrics, for example MAPE. Including the metrics for the per time series metrics would seem to be more consistent, and given they're being calculated for the aggregate metrics using numpy in gluonts.evaluation.Evaluator().get_aggregate_metrics() , it wouldn't be computationally expensive.
Whenever an instance of
gluonts.evaluation.Evaluator()is called with time series and forecast iterators, aggregate and per time series metrics are returned. A simple example of this is in the tutorial: https://ts.gluon.ai/stable/tutorials/forecasting/extended_tutorial.html?highlight=evaluator#Compute-metrics.Some of the metrics are included in the aggregated metrics, but not the per time series metrics, for example
MAPE. Including the metrics for the per time series metrics would seem to be more consistent, and given they're being calculated for the aggregate metrics using numpy ingluonts.evaluation.Evaluator().get_aggregate_metrics(), it wouldn't be computationally expensive.