I’m trying to assemble a process that can take sensor readings and categorise, from the peak shapes that appear, those that match the closest to a number of ‘indicators’ for a process problem. To achieve this, I have a data set of indicator peaks (see above) and ‘test’ peaks (see above) that I am running through the Functional Data Explorer platform. From this platform I’m applying the model to fit the ‘indicator’ peaks (leaving the test peaks as a validation set) and then saving the FPC scores that are generated.
What I want to do now is classify the ‘test’ peaks based on their FPC scores to those that most closely match the ‘indicator’ peaks (see an example visualisation below)– the problem I’m having is figuring out a suitable method to classify these peaks. I only have one category I’m aiming for which is ‘indicator’ – I don’t have another distinct group for ‘normal’ as that could be any number of shapes, forms etc. that would lead to a large training set. Can anyone make suggestion for how I can find the test peaks that most closely match?
I know clustering is possible, but would there be a way to cluster with a ‘control’ where they are clustered in comparison to the indicators?
If anyone as any other ideas in general for the approach I’m doing that would be great too.
Thanks!
