i have n3 input matrix and n1 output…n=695…when i use size(isoutlier(input)) i got 352…these values are important and should not be removed…also when i divide outliers and non-ooutliers…both of them show weak corelation to output…i use corr(inpu,out) in MATLAB…what can i do for handling both of these in MATLAB…input and outputs are scalar…is there any better architecture than legacy ANN than can handle?
1 Like
idk but how about robust regression?
https://mathworks.com/help/stats/robust-regression-reduce-outlier-effects.html