Table 6 Results of cross-validation of different RFs
From: Data-driven multinomial random forest: a new random forest variant with strong consistency
Task | Dataset | Algorithm | Performance | Optimal parameters |
|---|---|---|---|---|
Classification | Blogger | DMRF | 84.8 | \(B_{1} = 5,B_{2} = 8,p = 0.3,q = 0.8\) |
MRF(b) | 84 | \(B_{1} = 8,B_{2} = 5,q = 0.8\) | ||
BRF(b) | 84.2 | \(p_{1} = 0.05,p_{2} = 0.35,q = 0.8\) | ||
Denil14(b) | 83.2 | \(\lambda = 10,m = 500,q = 0.8\) | ||
Vertebral | DMRF | 84.76 | \(B_{1} = 2,B_{2} = 2,p = 0.5,q = 0.8\) | |
MRF(b) | 84.58 | \(B_{1} = 2,B_{2} = 10,q = 1 - 1/e\) | ||
BRF(b) | 84.6 | \(p_{1} = 0.05,p_{2} = 0.35,q = 1 - 1/e\) | ||
Denil14(b) | 84.45 | \(\lambda = 10,m = 200,q = 0.4\) | ||
House votes | DMRF | 96.26 | \(B_{1} = 2,B_{2} = 8,p = 0.3,q = 0.8\) | |
MRF(b) | 96.16 | \(B_{1} = 2,B_{2} = 2,q = 0.4\) | ||
BRF(b) | 96.14 | \(p_{1} = 0.05,p_{2} = 0.65,q = 1 - 1/e\) | ||
Denil14(b) | 95.85 | \(\lambda = 10,m = 50,q = 0.4\) | ||
Regression | ALE | DMRF | 0.0436 | \(B_{1} = 8,B_{2} = 10,p = 0.5,q = 0.4\) |
MRF(b) | 0.0443 | \(B_{1} = 8,B_{2} = 2,q = 1 - 1/e\) | ||
BRF(b) | 0.0451 | \(p_{1} = 0.01,p_{2} = 0.01,q = 0.4\) | ||
Denil14(b) | 0.0443 | \(\lambda = 5,m = 50,q = 0.4\) | ||
Real estate | DMRF | 51.54 | \(B_{1} = 5,B_{2} = 2,p = 0.3,q = 0.8\) | |
MRF(b) | 52.83 | \(B_{1} = 5,B_{2} = 2,q = 0.8\) | ||
BRF(b) | 52.49 | \(p_{1} = 0.01,p_{2} = 0.65,q = 0.8\) | ||
Denil14(b) | 58.56 | \(\lambda = 1,m = 50,q = 0.8\) | ||
Flare | DMRF | 0.5099 | \(B_{1} = 5,B_{2} = 5,p = 0.5,q = 0.2\) | |
MRF(b) | 0.5227 | \(B_{1} = 2,B_{2} = 2,q = 0.2\) | ||
BRF(b) | 0.5102 | \(p_{1} = 0.01,p_{2} = 0.01,q_{n} = 0.2\) | ||
Denil14(b) | 0.5156 | \(\lambda = 1,m = 100,q = 0.2\) |