Hello,
I am trying to adopt the Morkowitz optimization to partition data into an arbitrary but known number of minimum absolute errors sets.
My objective is to formulate the problem in a way that will return the regression parameters (b0 and b1 for each set) as well as the x-values for the partitions for the single predictor single response case. Down the line I will attempt to extend the formulation to MIMO cases.
Does anyone know of some valuable readings or tutorials that might help me along my way? Any insight would be greatly appreciated. The ultimate goal is to develop polyhedra of chemical engineering process data for use in gain scheduling control systems - so something a bit different that the financial engineering I see a lot of.
Edit: There is an additional constraint that the sets must contain all data points there within. Meaning if my partition is between x=0 and x=5 then all data points between those two belong only to that partition.
Kind Regards,
Richard
