Read Every Relevant Help Article, Watched the Videos. Cannot get RM to Predict Future

SkyTrader

Hi

I have asked questions here and read all relevant help materials but I'm close to moving on from RM as I am beset by problems and there is not a fix. The latest issue is my Excel attributes suddenly stopped populating the time series drop down for any of my processes.

I had successful ARIMA models giving future forecasts using the Apply Forecast, that is until one day about two weeks ago they stopped giving them...

I've never been able to get the Random Forest to predict into the future using Apply Model at all.

Does anyone else have these future prediction problems?

Is there a solution?

I only ask as I have spent two weeks trying to get RM to work and it doesn't. If not then it renders the whole point of why I am using RM moot.

Cheers,

MartinLiebig

Hi SkyTrader,

attached is an example which uses the Daily APPL stock from the past few years, optimizes an ARIMA and forecasts the next 7 days. I hope this helps.

Best,

Martin

SkyTrader

Thanks very much Martin @mschmitz, very much appreciated.

I replicated everything including sub processes (using the wizard) but the model will not run because I get this warning which I've had before and not been able to resolve?:



Cheers again for the rmp,

SkyTrader

Thank you very much Martin, @mschmitz, I got your sample to work, I got rid of the error by setting error handling to ignore.

Quick question my Performance operator has "relative error" ticked but I don't see it in my results, i.e. there's not a Performance Vector window in Results, the only Result windows I have are Apply Forecast, Optimise Grid and Example Set?

Many thanks,

MartinLiebig

Hi,

you probably need to connect the per port of optimize to the right hand side results.

Best,

Martin

SkyTrader

D'oh!

hughesfleming68

Very generally, you are running into extrapolation issues. Not all machine learning operators extrapolate the way you might expect. I highly recommend using synthetic data like sine waves and work with those first until you understand the limitations of each individual prediction operator.