Many existing clustering methods rely on the experience and insight into the original data of the analyst to determine the final model structure featured by number of groups (C) either directly (i. e., the fuzzy ISODATA analysis) or indirectly (i. e., the TWINSPAN classification). Hence certain arbitrariness and subjectivity are inevitably involved in the result of analysis when the individuals consisting of the data are not significantly clustered and when the analyst has difficult to gain insight into the...