“…A large K may cause the amount of data in certain OC class to be very small, and then influence the subsequent training procedure. Thus, we should limit the size of OC clusters K. As to the method for clustering, since the Phase I clustering method is not the focus of this paper, some existing works (Zorriassatine, Tannock, & O'Brien, 2003;Chakraborti, Van der Laan, & Bakir, 2008;Zou, Tsung, & Liu, 2008) are suggested to achieve good clustering performance, instead.…”