Table 2. Comparison of test performance (AUC) and complexity (I: number of intervals, V: number of variables) for all ICS setups (lpICS and enICS, each with and without preselection), Decision Tree (DT), Naive Bayes, linear and nonlinear (RBF) SVM.
Datasets include Acute inflammation (‘Inflammation’ and ‘Nephritis’ labels), Breast Cancer Diagnosis, Cardiotocography (‘Cardio’), Chronic kidney disease (‘Kidney’) and Indian Liver Patient data (‘Liver’).
| Inflammation | Nephritis | Breast cancer | Cardio | Kidney | Liver | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AUC | I (V) | AUC | I (V) | AUC | I (V) | AUC | I (V) | AUC | I (V) | AUC | I (V) | |
| lpICS | 1 | 6 (3) | 1 | 8 (4) | 0.933 | 5 (2) | 0.945 | 11 (5) | 0.932 | 4 (2) | 0.677 | 60 (9) |
| lpICS-pre | 1 | 6 (3) | 1 | 4 (2) | 0.933 | 5 (2) | 0.927 | 96 (8) | 0.959 | 8 (3) | 0.688 | 20 (5) |
| enICS | 0.962 | 4 (2) | 1 | 4 (2) | 0.942 | 4 (2) | 0.959 | 14 (6) | 0.938 | 6 (3) | 0.685 | 6 (2) |
| enICS-pre | 0.955 | 4 (2) | 1 | 4 (2) | 0.942 | 4 (2) | 0.873 | 6 (3) | 0.941 | 8 (2) | 0.685 | 6 (2) |
| DT | 1 | / (4) | 1 | / (2) | 0.936 | / (3) | 0.935 | / (13) | 0.952 | / (4) | 0.659 | / (3) |
| Naive Bayes | 1 | NA | 1 | NA | 0.976 | NA | 0.938 | NA | 0.964 | NA | 0.720 | NA |
| SVM-lin | 1 | NA | 1 | NA | 0.993 | NA | 0.957 | NA | 0.948 | NA | 0.696 | NA |
| SVM-rbf | 1 | NA | 1 | NA | 0.994 | NA | 0.957 | NA | 0.985 | NA | 0.706 | NA |