Table 4 Classification accuracy of naïve Bayes, k-NN, and SVM over the ten datasets
Datasets | Dimensions | Classifiers | ||
|---|---|---|---|---|
Naïve Bayes | k -NN | SVM | ||
Abalone | Original (8) | 22.10% | 26.01% (k = 9) | 25.19% (γ = 0.5) |
+2D (10) | 22.84% | 25.00% (k = 8) | 25.74% (γ = 0.5) | |
2D | 16.50% | 19.92% (k = 15) | 19.88% (γ = 0.5) | |
Balance scale | Original (4) | 86.70% | 88.46% (k = 14) | 90.54% (γ = 0.1) |
+2D (6) | 88.14% | 92.63% (k = 14) | 90.87% (γ = 0.1) | |
2D | 50.96% | 43.59% (k = 14) | 49.68% (γ = 0.1) | |
Corel | Original (89) | 14.34% | 16.50% (k = 11) | 20.30% (γ = 0) |
+2D (91) | 14.47% | 5.88% (k = 1) | 5.79% (γ = 0) | |
2D | 3.24% | 2.10% (k = 13) | 2.27% (γ = 0) | |
German | Original (20) | 72.97% | 69.00% (k = 6) | 69.97% (γ = 0) |
+2D (22) | 73.07% | 68.80% (k = 14) | 69.97% (γ = 0) | |
2D | 69.47% | 69.80% (k = 12) | 69.97% (γ = 0) | |
Hayes-Roth | Original (5) | 45.04% | 46.97% (k = 10) | 38.93% (γ = 0) |
+2D (7) | 35.11% | 45.45% (k = 10) | 40.46% (γ = 0) | |
2D | 31.30% | 46.97% (k = 2) | 36.64% (γ = 0) | |
Ionosphere | Original (34) | 81.71% | 86.29% (k = 7) | 92.57% (γ = 0) |
+2D (36) | 80.86% | 90.29% (k = 5) | 93.14% (γ = 0) | |
2D | 72% | 84.57% (k = 2) | 78.29% (γ = 0) | |
Iris | Original (4) | 95.30% | 96.00% (k = 8) | 96.64% (γ = 1) |
+2D (6) | 94.63% | 94.67% (k = 5) | 95.97% (γ = 1) | |
2D | 81.88% | 85.33% (k = 11) | 85.91% (γ = 1) | |
Optical recognition of handwritten digits | Original (64) | 91.35% | 98.43% (k = 3) | 73.13% (γ = 0) |
+2D (66) | 91.37% | 98.01% (k = 1) | 57.73% (γ = 0) | |
2D | 32.37% | 31.71% (k = 13) | 31.11% (γ = 0) | |
Teaching assistant evaluation | Original (5) | 52% | 64.00% (k = 1) | 62% (γ = 1) |
+2D (7) | 53.33% | 70.67% (k = 1) | 63.33% (γ = 1) | |
2D | 38% | 68.00% (k = 1) | 58.67% (γ = 1) | |
Tic-Tac-Toe Endgame | Original (9) | 71.06% | 81.84% (k = 5) | 91.01% (γ = 0.3) |
+2D (11) | 78.16% | 85.39% (k = 3) | 93.10% (γ = 0.3) | |
2D | 77.95% | 94.78% (k = 5) | 71.47% (γ = 0.3) | |