data extension

Hello Alphien,

I hope you’re doing well and wish you happy holidays.

We are working on new strategies using Machine learning techniques. But we are facing a bit of a problem regarding the data available. We need an extending window for training and testing a solid model.

we actually need historical data from the periods before 2003 and/or after 2016, I want to know how we can get this data extension and if there is a limit its length ( 2 years/5 years /10 years).

Thanks in advance for your replies.


You won’t be able to get data before 2003 or after 2016. The data youre allowed to use should fall within the 2003::2016 universe. May I know what issues you are facing regarding the current data?


As the first weights have to be generated before / on 02/01/2004 (1 year calibration/training), the concern is that the model we are trying to implement is trained over a period longer than one year ! 


In this case I suggest that you start with an initial period of 1 year of training, and then apply an expanding window.

Hope this clears it up

Hello Yongcheng,

Happy new year to you and Aphien team,

Thanks for your reply and explantation, it’s clear for me,