AutoSeries Challenge :: Feedback Phase Forum

Go back to competition Back to thread list Post in this thread

> primary_id

In description we have "primary_id: a list of column names, identifying uniquely the time series." but data in "starting_kit" has some rows with same primary_id+primary_timestamp and different target, but both public datasets are unique as it should be.
So question is: can we assume that primary_id+primary_timestamp is unique and there is mistake in starter_kit's dataset?

Posted by: rekcahd @ Dec. 6, 2019, 11:20 a.m.

i think the demo has the problem and you can use new data in file 14.

Posted by: Morxrc @ Dec. 6, 2019, 12:36 p.m.

I hope so, but it's better to get confirmation from challenge team

Posted by: rekcahd @ Dec. 6, 2019, 2:54 p.m.

I've noticed that you rank very high, As a newbie in automl , i found that The number of updates has a huge impact on the results, Is there a good strategy to auto-choose it,I would appreciate it if you could help me with this problem。

Posted by: Morxrc @ Dec. 8, 2019, 4:28 p.m.

Dear participants,

We confirm to you that the dataset 14 has been updated and for all datasets, timestamp and primary_ids should be able to identify uniquely.

Sorry for the very late reply. Should you have any further question, feel free to reach us by autoseries2020@4paradigm.com

Best regards,
AutoSeries Organizers

Posted by: Zhen @ Dec. 19, 2019, 9:53 a.m.
Post in this thread