Ensembling Methods for Selecting a Splitting Attribute in Decision Trees Learning Algorithms
Table 2: Effectiveness of attribute selection methods
Methods for selecting a splitting attribute | Classification accuracy on the task: | |||||||||||
1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | |
Correlation-based | 0.56 | 0.41 | 0.28 | 0.84 | 0.96 | 0.75 | 0.93 | 0.2 | 0.69 | 0.57 | 0.72 | 0.37 |
Chi-Squared Score | 0.51 | 0.39 | 0.33 | 0.58 | 0.76 | 0.7 | 0.81 | 0.46 | 0.76 | 0.68 | 0.79 | 0.25 |
Variance ratio | 0.45 | 0.42 | 0.41 | 0.88 | 0.95 | 0.64 | 0.88 | 0.73 | 0.70 | 0.68 | 0.8 | 0.31 |
Fast correlation-based filter | 0.41 | 0.33 | 0.16 | 0.86 | 0.65 | 0.71 | 0.88 | 0.2 | 0.73 | 0.54 | 0.75 | 0.21 |
Fisher score | 0.54 | 0.38 | 0.27 | 0.87 | 0.76 | 0.75 | 0.89 | 0.35 | 0.74 | 0.75 | 0.8 | 0.40 |
Mean absolute difference | 0.49 | 0.53 | 0.24 | 0.88 | 0.92 | 0.67 | 0.95 | 0.82 | 0.59 | 0.82 | 0.74 | 0.23 |
Variance | 0.49 | 0.55 | 0.24 | 0.88 | 0.92 | 0.68 | 0.95 | 0.81 | 0.59 | 0.82 | 0.74 | 0.23 |
ReliefF | 0.63 | 0.56 | 0.24 | 0.85 | 0.95 | 0.65 | 0.96 | 0.86 | 0.69 | 0.83 | 0.8 | 0.24 |
Separation Measure | 0.69 | 0.55 | 0.75 | 0.88 | 0.95 | 0.73 | 0.9 | 0.93 | 0.77 | 0.83 | 0.79 | 0.43 |