Assume I have 100 time series with different start and end dates but the same frequency hence they mostly have different lengths.Each time series is stored as a form of dataframe. They all look like the following:
date item_id target
2020-01-10 'A' 5
2020-01-11 'A' 6
2020-01-12 'A' 7
2020-01-13 'A' 8
2020-01-14 'A' 9
The second time series is:
date item_id target
2019-01-10 'B' 1
2019-01-11 'B' 2
2019-01-12 'B' 3
2019-01-13 'B' 4
I concatenate the time series to a big data frame and use the from_long method to create a long data frame.
dataset = PandasDataset.from_long_dataframe(long_df, time_col='date', target_col='target', item_col='item_id')
However, this gives me the following error while training:
AssertionError: Dataframe index is not uniformly spaced. If your data frame contains data from multiple series in the same column ("long" format), consider constructing the dataset with PandasDataset.from_long_dataframe instead
Can someone explain if DeepAR handles time series with different lengths (start/end date)? If No, how can I make it work when I have such a situation, and if yes, can someone explain how to to solve the above error?
