Table 5 The comparison of the performance of the four clustering algorithms over different matrix transformation methods by various indices (Dunn, Davies–Bouldin, eta-squared and omega-squared) and correlations between MSt and MSb of clusters
Transformation methods | Clustering algo. | Dunn mean (SD) | Davies-Bouldin mean (SD) | Eta-squared mean (SD) | Omega-squared mean (SD) | Cor (MSt, MSb) |
|---|---|---|---|---|---|---|
Lower triangle | RCKmeans | 0.2 (0.24) | 8.06 (4.55) | 0.32 (0.18) | 0.28 (0.16) | 0.82 |
RCDTC | 0.23 (0.24) | 6.73 (3.63) | 0.4 (0.14) | 0.35 (0.14) | 0.9 | |
RCPCA | 0.18 (0.15) | 7.22 (4.53) | 0.45 (0.09) | 0.33 (0.12) | 0.83 | |
RCPCA_RBF | 0.14 (0.12) | 8.84 (5.33) | 0.32 (0.11) | 0.26 (0.12) | 0.78 | |
Upper triangle | RCKmeans | 0.17 (0.29) | 8.06 (4.55) | 0.32 (0.18) | 0.27 (0.17) | 0.82 |
RCDTC | 0.21 (0.24) | 6.58 (3.2) | 0.39 (0.14) | 0.34 (0.15) | 0.89 | |
RCPCA | 0.18 (0.18) | 7.02 (3.46) | 0.45 (0.08) | 0.34 (0.11) | 0.82 | |
RCPCA_RBF | 0.14 (0.14) | 8.66 (5.02) | 0.33 (0.1) | 0.26 (0.11) | 0.78 | |
Average | RCKmeans | 0.2 (0.22) | 8.31 (5.32) | 0.34 (0.18) | 0.29 (0.16) | 0.82 |
RCDTC | 0.23 (0.24) | 6.61 (3.26) | 0.41 (0.14) | 0.35 (0.15) | 0.87 | |
RCPCA | 0.17 (0.13) | 7.24 (4.02) | 0.44 (0.09) | 0.33 (0.12) | 0.81 | |
RCPCA_RBF | 0.14 (0.12) | 8.54 (4.64) | 0.33 (0.11) | 0.27 (0.11) | 0.78 | |
Reciprocal | RCKmeans | 0.21 (0.24) | 7.5 (3.57) | 0.33 (0.18) | 0.29 (0.17) | 0.82 |
RCDTC | 0.22 (0.24) | 6.8 (3.51) | 0.4 (0.14) | 0.35 (0.15) | 0.91 | |
RCPCA | 0.17 (0.13) | 6.83 (3.8) | 0.45 (0.08) | 0.34 (0.11) | 0.82 | |
RCPCA_RBF | 0.14 (0.11) | 8.74 (5.07) | 0.32 (0.11) | 0.26 (0.11) | 0.79 | |
Non-reciprocal | RCKmeans | 0.2 (0.22) | 7.8 (4.05) | 0.33 (0.17) | 0.29 (0.16) | 0.81 |
RCDTC | 0.23 (0.24) | 6.68 (3.29) | 0.4 (0.14) | 0.35 (0.15) | 0.91 | |
RCPCA | 0.19 (0.15) | 6.87 (3.57) | 0.46 (0.07) | 0.35 (0.11) | 0.81 | |
RCPCA_RBF | 0.14 (0.12) | 8.96 (5.22) | 0.32 (0.11) | 0.26 (0.12) | 0.78 |