Table 4

Performance measures of data mining algorithm at different levels of significance over Antibodies dataset
SIGNIFICANCE p < 5 x 10-8 p < 5 x 10-7 p < 5 x 10-6 p < 5 x 10-5
Algorithm
    Acc.
    Sp
    Sn
    AUC
    Acc.
    Sp
    Sn
    AUC
    Acc.
    Sp
    Sn
    AUC
    Acc
    Sp
    Sn
    AUC
    Avg.
R. Forest 90.0 93.0 90.0 0.96 90.0 91.0 90.0 0.97 92.0 94.0 92.0 0.96 94.0 96.0 94.0 0.97 93.3
Bayes Net 88.0 92.0 88.0 0.96 88.0 91.0 88.0 0.96 94.0 95.0 94.0 0.95 92.0 95.0 92.0 0.96 92.5
Naïve Bayes 88.0 94.0 88.0 0.96 88.0 94.0 88.0 0.96 88.0 94.0 88.0 0.96 88.0 94.0 88.0 0.96 91.5
SVM 80.0 86.6 80.0 0.86 86.0 89.9 86.0 0.89 94.0 96.6 97.0 0.95 96.0 96.9 96.0 0.96 90.7
MLP 80.0 89.8 80.0 0.91 86.0 89.9 86.0 0.96 94.0 96.6 94.0 0.99 dnf dnf dnf dnf 90.2*
SLR 84.0 91.6 84.0 0.89 86.0 83.2 86.0 0.92 90.0 93.5 90.0 0.97 92.0 95.0 92.0 0.96 90.1
KNN 82.0 90.7 82.0 0.92 84.0 88.7 84.0 0.94 86.0 91.2 86.0 0.95 92.0 96.4 92.0 0.95 89.4
Logistic R. 72.0 85.3 72.0 0.92 84.0 90.1 84.0 0.93 92.0 96.4 92.0 0.98 90.0 96.1 90.0 0.98 89.1
M5P 80.0 91.5 80.0 0.92 76.0 87.4 76.0 0.90 78.0 89.4 78.0 0.91 74.0 85.4 74.0 0.89 83.2
Hyper Pipes 64.0 83.6 64.0 0.90 72.0 84.9 72.0 0.90 80.0 87.5 80.0 0.92 80.0 87.1 80.0 0.93 81.3
K star 88.0 93.4 88.0 0.94 94.0 97.2 94.0 0.95 82.0 91.8 82.0 0.93 20.0 90.2 20.8 0.68 80.7
J48 80.0 92.5 80.0 0.86 72.0 87.0 72.0 0.87 70.0 87.6 70.0 0.79 64.0 86.1 64.0 0.77 78.4
ASC 82.0 91.7 82.0 0.87 72.0 82.9 72.0 0.82 70.0 87.8 70.0 0.76 64.0 88.5 64.0 0.75 77.9
Random Tree 72.0 90.3 72.0 0.81 64.0 82.1 64.0 0.73 68.0 87.7 68.0 0.78 74.0 89.7 74.0 0.82 76.2
VFI 72.0 88.5 72.0 0.86 64.0 91.9 64.0 0.85 58.0 94.7 58.0 0.86 52.0 94.5 52.0 0.89 75.5
LDA 68.0 84.5 68.0 0.88 40.0 81.1 40.0 0.71 42.0 89.7 48.8 0.54 20.0 88.4 25.0 0.58 60.4
K means 46.0 68.7 46.0 0.57 46.0 68.7 46.0 0.57 40.0 68.1 40.0 0.54 40.0 68.1 40.0 0.54 52.5

Acc: Accuracy, Sp: Specificity, Sn: Sensitivity, AUC: Area under ROC curve, Avg: Average score in % for each algorithms, dnf: Did not Finish”, * denotes Avg. from 3 significance levels. Measures >90% are marked in bold.

Kukreja et al.

Kukreja et al. BMC Bioinformatics 2012 13:139   doi:10.1186/1471-2105-13-139

Open Data