In general, other algorithms, except R-SEL, reported slightly worse but still considerable SSEs and SD validity indices than
CSI. For certain data sets, however, some algorithms did not seem to work well in terms of SD validity, e.g. SCS, KKZ and KR with Abalone data set and KKZ with Segmentation data set. Note that, in terms of class accuracy, the CSI algorithm
reported the best performance for all UCI data sets. This result ascertained that CSI can not only find better cluster structures, but also the discovered clusters are more analogous to the intrinsic class structures.