Table 5. Performance results for the top 5 WEKA models, from the ensemble selected for Shakespeare authorship attribution based on performance against Fletcher, Jonson and Middleton (see Materials and Methods section).
CM1 Score | Welch's t-test | |||||
Method | MCC | Spec. | Sens. | MCC | Spec. | Sens. |
SMO | 0.987 | 1.000 | 0.980 | 0.960 | 1.000 | 0.940 |
IB1 | 0.955 | 0.987 | 1.000 | 0.863 | 1.000 | 0.800 |
IBk | 0.955 | 0.987 | 1.000 | 0.863 | 1.000 | 0.800 |
MultilayerPerceptron | 0.934 | 0.933 | 0.920 | 0.930 | 0.993 | 0.920 |
Logistic | 0.923 | 0.960 | 0.980 | 0.916 | 0.967 | 0.960 |
Performance is evaluated in terms of Matthews' correlation coefficient, specificity (precision) and sensitivity (recall) for marker words selected by both CM1 and Welch's -test scores.