Alteryx Designer Desktop Discussions

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

Time Series Forecasting by Multiple Fields

NakulMalisetty
6 - Meteoroid

Hi Everyone,

 

I am building a forecast using the Time Series time series forecasting model. I tried with ARIMA which only allows to forecast one numeric field with date. I have multiple fields I need to forecast - is there a way I can add multiple fields so I could forecast for all the fields at one time?

 

Attached excel has the sample data from 2021- 2023(Nov) and highlighted data from Dec-2023 to the end of next year is to be forecasted in Alteryx. 

 

Thank you for the support in advance.

 

Thanks

Nakul

10 REPLIES 10
alexnajm
16 - Nebula
16 - Nebula

You cannot natively in the Time Series tool - you'd have to either do a new model for each column you want to forecast, or create some sort of batch macro.

NakulMalisetty
6 - Meteoroid

Hi Alex,

 

Thank you for the response. Coming to macro, I am trying with TS Model Factory.

Could you elaborate or share a sample to do a new model for each column?

 

Thanks

Nakul

alexnajm
16 - Nebula
16 - Nebula

Hi Nakul,

 

TS Model Factory will allow for multiple models based on a grouping field, but that doesn't sound like what you are trying to accomplish - it sounds like you are trying to affect the response variable. For example, it sounds like you are trying to forecast unit cost AND units. This would not be a case for the Factoty tools in that scenario.

 

I can try to help further once there is more clarity on the ask!

NakulMalisetty
6 - Meteoroid

Hi Alex,

 

I got few suggestions to try TS Model Factory to achieve what I need and I am still trying it. If this model doesn't help then there is no point of working on this model.

As the attached sample data, my requirement is, I have a date field, two numeric fields and few other fields. Now I need to forecast all the fields till the end of next year. I tried and ran out of ideas, it will be helpful if you provide a sample with the attached data in achieving this.

 

Thank you for the support.

 

Regards

Nakul

alexnajm
16 - Nebula
16 - Nebula

Hi Nakul,

 

I am sorry but your requirements are not fully clear, and I won't be able to help without clarity on more things.

  • What field frequency are you using (monthly, daily, etc.)?
  • If you are doing ARIMA, are there covariates to be aware of?
  • Why you are using the TS Factory tools? The only reason you'd use it is if you wanted to make a model per unique value of a column, i.e. Region (East, West, etc.). So what are you assuming your grouping column is?

 

You can include your workflow by clicking Options > Export Workflow so we could use it as the basis.

NakulMalisetty
6 - Meteoroid

Hi Alex,

 

I am new to Alteryx and randomly trying all the forecasting models to achieve this.

I attached the workflow for reference.

  • Date frequency is on daily basis however, the output I am trying to get is on 'Monthly'

Thanks

Nakul

 

alexnajm
16 - Nebula
16 - Nebula

Hi Nakul,

 

Apologies but I cannot help further without more information - here's what I could accomplish based on what you've described so far and the new data you have provided.

 

-Alex

NakulMalisetty
6 - Meteoroid

Hi Alex,

 

Thank you for the workflow! I already tried this way, here we are able to forecast one numeric field based on date however, I need all the other fields to be forecasted as well. Is there a way we can achieve this in Alteryx?

 

Regards

Nakul

alexnajm
16 - Nebula
16 - Nebula

Hi Nakul,

 

You only had one viable numeric field in there to calculate so I did it based off of that. If you want to repeat for others, you can repeat a similar process in that container with the Summarize, TS Filler, etc. These would just be multiple containers in the same workflow. I can't invest time creating a batch macro solution until more time is put into having data to do this on - I also think repeating the containers for each metric is absolutely viable since you might want to treat each one differently in some way.

 

-Alex

Labels