Table 8. Comparative performance analysis of the rule-based classifiers on Dataset 3, respectively (at 4-fold CVs repeating for 10 times); where bold font denotes the highest value for each column.
| Rule-based classifier | Average sensitivity[%] (s.d.) | Average specificity[%] (s.d.) | Average accuracy[%] (s.d.) | Average MCC (s.d.) |
|---|---|---|---|---|
| Proposed | 90.56 (2.68) | 86.67 (2.87) | 88.61 (2.43) | 0.77 (0.048) |
| ConjunctiveRule | 70.56 (2.68) | 90.56 (6.95) | 80.56 (2.27) | 0.63 (0.062) |
| DecisionTable | 84.44 (3.51) | 82.78 (4.86) | 83.61 (0.88) | 0.68 (0.019) |
| JRip | 75.56 (2.87) | 92.78 (2.68) | 84.16 (1.34) | 0.69 (0.027) |
| OneR | 76.67 (7.31) | 88.33 (4.86) | 82.50 (1.33) | 0.66 (0.014) |
| PART | 76.11 (13.11) | 94.44 (0.00)* | 85.28 (6.55) | 0.72 (0.113) |
| Ridor | 83.33 (4.54) | 80.00 (9.51) | 81.67 (2.68) | 0.64 (0.048) |
* This standard deviation of specificity is coming to be zero. On investigation, we have identified a particular datapoint belonging to normal class in Dataset 3 for which the “PART” classifier as well as the other classifiers including the proposed one are producing always false positive result.