Adding Seasonality month in Time series function

[Dhilip_Kumar]

Hi,

I have used ARIMA and ETS tools to forecast my future data. Now I want to add few things on that workflow, we received Jan and Jul month high sales rather than other months so How can I add this instruction into ARIMA and ETS tool function. I have attached my workflow for your reference please have a look at it and advise me.

Also, I am looking for same results while forecasting daily, we received high sales on 1st and 15th day of each month.

[aatalai]

@Dhilip_Kumar  take a look into this

1) Look at forecasting into the future containers - set  number of periods to forecast as to how far into the future you want to go ot

2) apologies missed and added

3) what would you like to output to excel

[Dhilip_Kumar]

@aatalai I apologies for multiple questions: 

1. Is that any specific reason to use ETS method for Reporting and ARIMA for Invoice. Its just curious to know that and would be helpful in future.

2. I could see two containers one for Reporting and another one for Invoice. what would be the solution if we have multiple categories.

3. I need the projection values on excel file instead of viewing it on Alteryx app.

[aatalai]

@Dhilip_Kumar no need to apologies that is how we learn 

1) they are different models and went for the one with lowest variance 

2) what do you mean? As it would mean most likely a model per category  and if you look at the training here https://community.alteryx.com/t5/Interactive-Lessons/Selecting-and-Scaling-Models/ta-p/610381 it talks about the factory tools that can help with scaling. 

3) look at the outputs from the forecast tool and output them to excel similar to how you would export data normally.

Let me know if you have any further qs/clarification

[Dhilip_Kumar]

@aatalai 

I am still little confusing on the workflow. You used TS Compare tool after using the ETS method for first 50 rows and ARIMA method for last 10 rows. Could you please explain why we should use this comparison tool and suggest that we repeat the ETS/ARIMA method for forecasting the future with the same set of data.

1. You mentioned low variance behind the reason for elect ETS or ARIMA tool to forecast, So if my understanding is correct we will need to use TS Compare tool to analysis which method is perfectly match for this category and then we will use the same to forecast. correct? If so, kindly confirm how could we do these comparison for each categories on TS Model Factory.

2. what would be preferable method to use this for forecasting, Since how the users will elect the model ETS or ARIMA after published on server.

[aatalai]

@Dhilip_Kumar happy to help but do watch the data science learning path https://community.alteryx.com/t5/Learning-Paths/Data-Science-Learning-Path/ta-p/504157#done

as this will help you with your understanding.

Yes you train your model with the first x rows (I choose 50 - 5/6) and then compare it with the last 190 to see how your model perfomrs with unseend ata - does that help clarify things?

1) I haven't used the factory tools myself, but assume that you take the O output from the TS model factory and plug it straight into the TS compare tool similar to how it has been done in the previous flows shared

2) You would only choose the object output of the model you have chosen as developer onto the server and set up forecast based on that - i.e for each model you will not be uploading both ETS and ARIMA, only one. And then based on the model (reporting/invoicing) the user wants to choose that is autos elected - i.e the final user wouldn't need to choose ETS/ARIMA just if it is invoicing or reporting. 

Let me know if you need any further guidance

[Dhilip_Kumar]

@aatalai I appreciate all your help. I will go through all these learning videos and try the same with your inputs. I will update you.