Fitted Values chart - how to get values?

[Dima1]

Hi All

Could you pls advise how to get fitted values based on prediction of arima/ets model for input data? I dont need prediction, but only fitted using model. Thanks!

[NeilR · Accepted solution]

The following seems to be working for me for the ETS tool:

1. hook up an R tool to the O output anchor of the ETS tool
2. paste in the following R code into the R tool

the.model <- read.Alteryx("#1")
mod.obj <- unserializeObject(as.character(the.model$Object[1]))
model.fitted <- fitted(mod.obj)
write.Alteryx(data.frame(Fitted=as.matrix(model.fitted)), 1)

[krishnapandey]

Hi Neil,

I am trying to use same code for getting fitted values for ARIMA model ARIMA(1,0,1)(1,0,1)12 but i am getting error "Error: R (95): Error in array(x, c(length(x), 1L), if (!is.null(names(x))) list(names(x), :
".  any thoughts?

Thanks,

Kris

[NeilR]

It's working for me. What Alteryx version are you on? Can you include the workflow and data that's giving you the error?

[krishnapandey]

Thanks Neil! It is working for me now.

I was using version 11.3 and had problems, upgraded to 2018.2 and it is working. 

Thanks

[cwkoops]

Hi,

I'm using the TS Factory tools to get predictions on groups of data.  I'm interested in the next forecasted point, as well as the fitted values - similar to the need in the original post.  I am leveraging the R code Neil provided, however it doesn't appear to be giving fitted values for anything beyond the first group.  What would needs to be added / changed with the R code so the fitted values are listed by group? See attached workflow.  Thanks!

[NeilR]

@cwkoops could you attach a package (Options->Export Workflow) rather than a workflow? The TS Factory tools aren't getting included.

[cwkoops]

Hi Neil,

I attached the .yxzp.  Below is an example of the desired output from the R tool, if I allow multiple groups to flow into the TS Factory tools.  Thanks for taking a look. 

[NeilR]

The TS Model Factory tool outputs each group as a separate model object in a new row. The second line of the original R code is only pulling from row 1:

mod.obj <- unserializeObject(as.character(the.model$Object[1]))

Changing that 1 to a 2 gets you the second group. You'll likely want to iterate across each row (for loop or something along those lines) to get a more robust solution.

[cwkoops]

Thanks Neil.  I was playing around with those values, but nothing was obvious to me.  This seems to work with the for loop you mentioned.  

the.model <- read.Alteryx("#1", mode = "data.frame")
for(i in 1:nrow(the.model)){
mod.obj <- unserializeObject(as.character(the.model$Object[i]))
model.fitted <- fitted(mod.obj)
write.Alteryx(data.frame(Fitted=as.matrix(model.fitted)), 1)
}

So now I have the correct # of rows, but the 2nd column with the group identifier would be nice, but possibly not necessary as I could just append it to the original input data stream...if the # of rows is the same.  Any thoughts there?