I was working in the MLTK, very new to it and exploring. I was working to establish a few searches where I will fit a algorithm and then apply it to identify if any values out of a set boundary and then alert on that. I have two question from this.
Is this a valid use case or not so much?
I have a predicted value after my fit but, its too close to my actual values so I was thinking of doing something like(+ or - depending on need):
eval bound = (predictedavg - (stdev * 3))
Would it be more beneficial to calculate this in the fit search or when applying the model?