Community Spring Cleaning week is here! Join your fellow Maveryx in digging through your old posts and marking comments on them as solved. Learn more here!

Alteryx Designer Desktop Discussions

Find answers, ask questions, and share expertise about Alteryx Designer Desktop and Intelligence Suite.
SOLVED

Questions about times series

Gualigee
8 - Asteroid

Gualigee_0-1651052745769.png

why did I have the same prediction for each month?

I used factory tools. It shows a different results. Since factory tools cannot edit the parameters so I selected auto in the ETS tool. 

Attached are the workflow and input data. 

 

Thanks in advance. 

 

6 REPLIES 6
DEllis
7 - Meteor

You have the seasonal type set to 'Auto' in the model tab - I'm not sure what criteria gets applied in auto mode, but it is clearly not using seasonality in this case if you get the same estimate for each month. If you change the mode to additive or multiplicative, you get different results for each month, though they are still different from the factory. You'd have to dig into the results anchors to see exactly what parameters it is using for each model.  Based on a quick glance the factory is using ETS(m,n,a) and the ETS tool is using ETS(A,Ad,A). I don't know enough about the underlying model to tell you how/if those parameters are different.

Gualigee
8 - Asteroid

@DEllis Thank you for your reply. I set ETS tool to (m,n,a) or (m,n,m), the difference between ETS tool and Factory tool is more than 10%. but at least each month is different. My understanding is ETS tool's auto recommendation uses one of the parameter combinations with lowest AIC,  

Gualigee
8 - Asteroid

I found another post

https://community.alteryx.com/t5/Alteryx-Designer-Discussions/Time-Series-Analysis-Not-Predicting-Su...

 

I think the problem is for each individual ETS tool, the input data is too small to predict seasonality if you want to use the auto feature. 

DEllis
7 - Meteor

@Gualigee That would certainly make sense. There is a risk of overfitting if you try to apply seasonality with a limited number of data points for each seasonal period. 

LiuZhang
9 - Comet

Screenshot 2022-04-28 184520.png

Actually, you forgot to sort the data.

 

Gualigee
8 - Asteroid

Thank you, @LiuZhang. I never thought there should be a sorting step. 

Labels