How to use conditionals in composed pipelines

My use case is the following: i want to use an ARIMA model to do timeseries forcasting, but i dont want outliers to “pollute” the forecast quality. Those outliers are usally bad sampled data deriving from bad sensor settings or maintenance and are in generally easy to figure out (sudden drops in values, or so)

So the idea would be to use pipe an anomaly detector before the arima model and when an anomaly is detected then the model.learn function is not called.

Sure enough i could implement those 2 models as independent models and an if/else statement in betwee, but i was wondering if it is possible to have all of it into a single pipeline.

Like:

StandardScaler --> AnomaliesDetector --> Thresholder -->  If(isNotAnomaly) --------- true -----> ArimaModel --> done
                                                               |----- false ------> skip learning

Is this possible in river ?

[MaxHalford]

I’m on my phone so I can’t give you details instructions. It actually happens with pip, it’s just that you have to tell pip to install from sources on GitHub. There are I strictions in the installation section in the docs.

[MaxHalford]

River doesn’t provide any such functionality, but maybe you’ll find what you need in pandas.