Skip to main content

Table 2 Performance of established classification techniques compared to bagged J48.

From: Enhancing navigation in biomedical databases by community voting and database-driven text classification

  Bagged
J48
Naïve
Bayes
SVM k-Nearest Neighbor Trees PART
Dataset    Polynomial kernel degree    J48 Random
Forest
 
    1 2 3 1 3    
CA 92.1 (5.0) 87.1 (8.0) 91.4 (5.7) 89.4 (6.6) 79.4 (5.9) • 66.9 (10.3) • 73.2 (9.2) • 88.7 (6.5) 81.7 (8.0) • 86.3 (8.5)
CVD 91.8 (4.6) 89.5 (7.2) 92.5 (4.7) 89.2 (5.1) 83.6 (6.5) • 56.9 (12.4) • 64.3 (14.3) • 90.5 (5.7) 83.9 (6.7) • 85.6 (6.8)
DM 96.4 (3.6) 83.3 (8.3) • 95.2 (3.8) 89.0 (5.2) • 78.4 (4.0) • 74.4 (6.9) • 75.9 (6.3) • 96.5 (4.5) 83.1 (6.8) • 96.2 (4.7)
APO 98.3 (3.1) 94.8 (4.8) 98.3 (2.6) 96.3 (4.0) 41.7 (22.1) • 24.2 (18.4) • 33.4 (18.6) • 98.1 (3.8) 86.9 (7.4) • 97.2 (3.7)
ANG 95.3 (4.2) 94.3 (4.5) 95.3 (4.1) 93.4 (4.4) 91.5 (7.0) 47.7 (18.3) • 48.5 (18.5) • 94.0 (5.4) 89.5 (5.7) 91.3 (6.1)
MI 94.5 (3.9) 91.0 (5.6) 94.5 (4.1) 89.9 (4.4) 77.1 (4.0) • 73.4 (4.9) • 71.9 (4.1) • 92.1 (5.0) 85.4 (7.1) • 92.2 (4.8)
BD 90.7 (5.3) 87.1 (6.2) 91.1 (5.4) 88.0 (5.6) 75.8 (4.5) • 69.7 (2.9) • 67.4 (1.3) • 86.7 (6.4) 82.3 (7.8) 83.4 (7.4) •
  1. F-Measure (in %) and standard deviation in parentheses. • statistically significant degradation. No statistically significant improvement was observed.