algorithm - K-Means for diagonal clusters -
i have 2 clusters lie along 2 lines on 3d surface:
i've tried simple kmeans algorithm yielded above separation. (big red dots means)
i have tried clustering soft k-means different variances each mean along 3 dimensions. however, failed model this, presumably cannot fit shape of diagonal gaussian.
is there algorithm can take account data diagonal? alternatively there way "rotate" data such soft k-means made work?
k-means not prepared correlations.
but these clusters reasonably gaussian me, should try gaussian mixture modeling instead.
Comments
Post a Comment