Cool approach, I will look into it further and try to fine tune it to the actual data set. Thanks for the pointers.