From: Learning accurate and interpretable models based on regularized random forests regression
Numbers after ± are standard deviation. SVR is support vector regression. | |||
---|---|---|---|
 | Random Forests | Our Approach | SVR |
Stockori Flowing Time | |||
R 2 | 0.54 ± 0.00 | 0.45 ± 0.05 | 0.28 ± 0.03 |
Number of Rules Selected | 66020 ± 187 | 348 ± 33 | NA |
Number of Features Used in a Rule | 8.8 ± 1.9 | 7.5± 1.74 | NA |
Number of Features Selected | 149 ± 0 | 135 ± 31 | 149 ± 0 |
Parkinson's Telemonitoring | |||
R 2 | 0.15 ± 0.02 | 0.06 ± 0.02 | 0.17 ± 0.02 |
Number of Rules Selected | 644789 ± 414 | 3796 ± 0 | NA |
Number of Features Used in a Rule | 9.72± 2.14 | 7.4 ± 1.86 | NA |
Number of Features Selected | 19 ± 0 | 19 ± 0 | 19 ± 0 |
Breast Cancer Wisconsin (Prognostic) | |||
R 2 | 0.04 ± 0.02 | -0.19 ± 0.16 | -0.04 ± 0.04 |
Number of Rules Selected | 43907 ± 58 | 126 ± 2 | NA |
Number of Features Used in a Rule | 7 ± 3 | 3 ± 1.49 | NA |
Number of Features Selected | 32 ± 0 | 31 ± 1 | 32 ± 0 |
Relative location of CT slices on axial axis | |||
R 2 | 0.92 ± 0.01 | 0.77 ± 0.09 | 0.26 ± 0.00 |
Number of Rules Selected | 172984 ± 143 | 901 ± 15 | NA |
Number of Features Used in a Rule | 12 ± 3.12 | 8 ± 2.53 | NA |
Number of Features Selected | 384 ± 0 | 20 ±± 5 | 384 ± 0 |
Seacoast | |||
R 2 | 0.64 ± 0.02 | 0.59 ± 0.10 | -0.19 ± 0.00 |
Number of Rules Selected | 120771 ± 161 | 385 ≤ 5 | NA |
Number of Features Used in a Rule | 14 ±± 3 | 6 ± 1.91 | NA |
Number of Features Selected | 16 ± 0 | 16 ± 0 | 16 ± 0 |
TCGA Glioblastoma multiforme | |||
R 2 | 0.04 ± 0.01 | -1.94 ± 0.67 | -0.09 ± 0.00 |
Number of Rules Selected | 53539 ± 31344 | 279 ± 6 | NA |
Number of Features Used in a Rule | 3 ± 2 | 2 ± 1 | NA |
Number of Features Selected | 12042 ± 0 | 2 ± 1 | 12042 ± 0 |