I have multiple measurements related to the object of interest (stationary). I want to identify the type of object.
I am computing a convex sum of the probability the type of object using the sum. sum(I) =weight* sum(i-1) + (1- weight)*measurement(i)
I am plotting the sum(belief) with respect to time for different classes in the scope in the attached image.
I want to know a method to optimize such I can assign the appropriate weights to the sequential sum such that fluctuation in the trajectory of sum is minimum.
Any related works or suggestions will be good. Is it possible to use ensemble trees?