community
cancel
Showing results for 
Search instead for 
Did you mean: 

Alteryx Designer Knowledge Base

Definitive answers from Designer experts.

Time Series Factory Tools vs Batch Macro Time Series

Alteryx
Alteryx
Created on
image1.jpgdcvsmarvel.fandom.com/wiki/Superman_vs._Hulk

 

 

 

The pre-requisites to reading this article are Performing Time Series Forecasting in Alteryx Designer written by @PamW, and Build Your First Batch Macro! written by @JoeM.

 

Imagine you have a data set with sales data for eight different wines for years 1992 to 2018 and you’d like to predict the next 6 years of sales for each of these. This can be performed by building a batch macro around the Time Series tools, or by using the Time Series (TS) Factory tools.

 

 

Factory Time Series

Batch Macro Time Series

Ease of use, tools just need configuring

Macro requires building, does not come pre-built

Shorter run time

Longer run time

Minimum record count: Annual – 2 records, Quarterly – 8 records, Monthly – 24 records [Note: you always need 2 years of data for each time unit]

No minimum historical record count, but fewer records produce a poor forecast with little to no change

No interactive reporting

No interactive reporting

Date aggregation functionality – e.g. monthly data and forecasting on a yearly basis, automatically aggregates months in the same year, but require a date type field

No date aggregation functionality

Automatically brings through group name for which you are forecasting (e.g. Wine type) 

Group name (e.g. Wine type) is not automatically appended to forecast output, must be configured in macro

ARIMA and ETS modeling

ARIMA and ETS modeling

 

 

 

Below is the workflow using the batch macro. The batch macro requires a control parameter (indicated by the “¿” input anchor), which in this case is the wine type; hence a list of wine types is passed to the macro, as well as the data set containing all the sales data.




clipboard_image_3.png

 

 

 

The workflow inside the macro can be seen below. The macro inputs all the wine data, then filters by the wine type from the wine type control parameter list. The sales data for each wine type is then used to build the time series model and forecast the next six years. The wine type is then appended to the forecasted years and output. This is done for each wine type, with the forecasting results being stacked on top of one another as the workflow output.

 

clipboard_image_4.png



The output of the workflow can be seen below, with each run of the macro being stacked onto one another. The configuration for each of the tools and the macro can be seen in the attached workflow.


clipboard_image_5.png



The alternative to using a batch macro to predict wine sales is the TS Factory time series tools, which can be downloaded via the Alteryx public gallery (TS Model Factory, TS Forecast Factory). The TS Model Factory tool is used to build the time series models and the TS Forecast Factory is used to forecast a user-defined number of periods. The workflow using these tools can be seen below.

 

clipboard_image_0.png

 

 

The TS Factory tools can build time series models and forecasts for multiple sets of historical data groups, without the need to separate them. All that is required is that within your data, exists a field to identify each set of historical data so it can be treated as a separate time series. In the wine example, this field is the Wine Type field, much like the control parameter for the batch macro. Then like the ordinary time series tools, a time series forecast is produced for the next 6 years based on the model built for each wine type. The output of this can be seen below:

 

clipboard_image_7.png



In conclusion, when working with multiple sets of historical data, there is more advantage in using the time series factory tools, especially if you are not familiar with building macros. Additionally, the factory time series are quicker to run and achieve the same result as the batch macro, with fewer tools and more functionality.

 

Attachments
Comments
Alteryx
Alteryx

This is great! Thank you for the information  😁