Descriptive statistics of the sizes of the regions and the number of the singleton clusters obtained by each used algorithm are presented in Table 4. As seen in Table 4, SLFCM and ALFCM algorithms tend to form regions with noticeable differences in their sizes (i.e. a small number of large regions and several small regions). These findings are consistent with the results reported by Rao & Srinivas (2006a). On the other hand, CLFCM, WFCM and SOFMFCM algorithms often form regions with sizes relatively close to each other. Also, they do not tend to form regions including one site (singletons). In general, singleton regions are not appropriate to perform RFFA and therefore the clustering algorithms that tend to form singleton clusters are not reasonable choices for regionalisation.
Main statistical characteristics of sizes of the regions formed by the hybrid algorithms
Algorithm . | Mean . | Std. dev. . | Max . | Min . | Singletons . | Total number of sites belonging to all clusters in m = 2.5 for c = 7 . |
---|---|---|---|---|---|---|
SLFCM | 9.595 | 5.978 | 26 | 1 | 1 | 65 |
CLFCM | 9.578 | 5.757 | 26 | 3 | 0 | 65 |
ALFCM | 9.560 | 5.799 | 26 | 1 | 3 | 65 |
WFCM | 9.567 | 5.044 | 26 | 3 | 0 | 65 |
SOFMFCM | 9.564 | 4.851 | 26 | 3 | 0 | 67 |
Algorithm . | Mean . | Std. dev. . | Max . | Min . | Singletons . | Total number of sites belonging to all clusters in m = 2.5 for c = 7 . |
---|---|---|---|---|---|---|
SLFCM | 9.595 | 5.978 | 26 | 1 | 1 | 65 |
CLFCM | 9.578 | 5.757 | 26 | 3 | 0 | 65 |
ALFCM | 9.560 | 5.799 | 26 | 1 | 3 | 65 |
WFCM | 9.567 | 5.044 | 26 | 3 | 0 | 65 |
SOFMFCM | 9.564 | 4.851 | 26 | 3 | 0 | 67 |