Hi,
I'm performing an ARIMA model using the ARIMA tool with two covariates. I specified the Mean of the covariates to use by the ARIMA tool for the covariates. To make forecasts, I'm using the TS Covariate Forecast tool.
For your information, the TS Covariate Forecast tool requires 2 input data streams. I made a first connection between the O output from my covariate ARIMA model and the R Input anchor and a second connection between my original dataset fed into my covariate ARIMA model and the L Input anchor of the TS Covariate Forecast tool. I wanted to forecast the next 6 periods. However, the TS Covariate Forecast tool does not only forecast the 6 next periods but also forecasts the next 43 periods. I mean I got a table with 43 records from the O anchor. For your information, I have a dataset of 43 observations. Could you please provide me any explanations of why I'm getting this number of records from the TS Covariate Forecast tool? Thank you!
The TS Covariate Forecast tool will forecast for each row input into the L anchor.
If you have 43 rows going into this that will explain the output.
If you have selected to just do 6 periods in the ARIMA tool, that should be all it will do on the report output there, but it's not a parameter that's passed through into the forecast tool itself.
Hi @JoeS,
Thank you for taking the time to give me a solution. I do select to do 6 periods in the ARIMA tool, and this is what I'm expecting to get. Otherwise, more than requested, I got 43 records from August 2022 to February 2026.
I think if you have selected 6 in there that will be what you see out of the ARIMA model tool in the R and I anchor.
The forecast tool has no knowledge of what you have selected in the ARIMA model, it will simply build a forecast for each of the rows you have going into the L anchor.
If you only want the first 6 out of the forecast tool, you can add a sample tool to just have August 2022 to February 2023.
Hi Joe,
I believe that the ARIMA model is making forecasts based on all the rows where I have historical values, not for each row that I have in the L anchor.
By the way, we don't have this problem with the TS Forecast tool which will generate only the number of forecasts expected based on what I indicated in the ARIMA tool. I got only this problem with the TS Covariate forecast tool.
Thank you so much for the time that you've provided!
Yeah, sorry it's making the forecasted values/amounts based on the historic values, but in terms of the number of future forecasts it makes it for each row you have into the L anchor.
I think with the normal forecast tool you aren't seeing an issue as the default is 6.
If you change the number on the forecast tool it should then change the future number of forecasts to what you have entered there and ignore what you have put into the original ARIMA tool
Hi @JoeS
I believe there is no way to configure the number of expected forecasts in the TS Forecast tool, as we can only do this in the ARIMA tool, which will be connected to the Forecast tool.
For the TS Covariate forecast tool, I looked at the R code available in the macro of this tool. I note that the problem I'm facing is related to how the number of expected forecasts is configured in the R code, as below. The number of the generated forecasts should equal the matrix of covariates which also should equal the number of records of my data stream.
Besides, this is how the number of expected forecasts is configured in the R code of the macro of the TS Forecast tool, where the number of forecasts could be adjusted according to the value indicated by the user in the ARIMA tool:
What do you think?
I am not sure if we are talking cross purposes here?
You can specify it in the configuration panel of the TS Forecast tool at the bottom:
And you are right, when it comes to the covariate your right, it's not a configuration option, it's based on the number of rows input.
Yes, I'm sorry I meant we don't have this option in the TS Covariate Forecast tool.
Thank you for your time!