Table 3

Performance measures of data mining algorithm at different levels of significance over Alzheimer’s dataset
SIGNIFICANCE p < 5 x 10-5 p < 5 x 10-4 p < 5 x 10-3 p < 5 x 10-2
Algorithm
    Acc.
    Sp
    Sn
    AUC
    Acc.
    Sp
    Sn
    AUC
    Acc.
    Sp
    Sn
    AUC
    Acc
    Sp
    Sn
    AUC
    Avg.
Naïve Bayes 100 100 100 1.00 91.3 82.0 100 0.96 91.3 82.0 100 0.96 86.5 91.0 84.0 0.94 93.4
Logistic. R 95.0 90.0 100 0.99 95.7 90.0 100 0.97 91.3 90.0 91.7 0.90 91.3 90.0 91.7 0.90 93.3
MLP 91.3 90.9 91.7 0.97 95.6 90.9 100 0.97 87.0 90.9 83.3 0.97 dnf dnf dnf dnf 92.7*
VFI 91.3 90.9 91.7 0.87 95.7 90.9 100 0.92 91.3 81.8 100 0.89 91.3 81.8 100 1.00 92.2
KNN 91.3 90.9 91.7 0.93 95.6 90.9 100 0.93 86.9 90.9 83.3 0.95 91.3 90.9 91.7 0.92 91.8
K-means 82.6 100 66.7 0.83 91.3 90.9 100 0.91 95.7 90.9 100 0.96 91.3 81.8 100 0.90 90.7
Hyper Pipes 91.3 81.8 100 0.98 95.7 90.9 100 0.97 91.3 81.8 100 0.95 73.9 81.8 66.7 0.90 89.7
SVM 87.0 90.9 83.3 0.87 95.7 90.9 100 0.95 82.6 81.8 83.3 0.83 87.0 81.8 91.7 0.87 88.0
Bayes Net 91.3 81.8 100 0.96 91.3 90.9 91.7 0.95 87.0 81.8 91.7 0.86 78.3 81.8 75.0 0.84 87.7
R. Forest 86.9 81.8 91.7 0.94 82.6 81.8 83.3 0.93 73.9 72.7 75.0 0.89 72.6 81.8 75.0 0.84 82.4
K star 95.7 90.9 100 0.98 91.3 90.9 91.7 0.94 78.2 81.8 75.0 0.86 56.5 18.2 91.7 0.64 81.5
SLR 86.9 81.8 91.7 0.96 73.9 72.7 75.0 0.82 60.9 63.6 58.3 0.80 52.2 54.5 50.0 0.69 71.8
Random Tree 78.3 72.7 83.3 0.78 60.9 54.5 66.7 0.61 73.9 63.6 83.3 0.74 73.9 81.8 66.7 0.74 71.7
ASC 73.9 63.6 83.3 0.61 68.9 63.6 58.3 0.56 73.9 81.8 66.7 0.75 78.2 63.9 91.7 0.61 70.0
J48 73.9 63.6 83.3 0.61 60.9 63.6 58.3 0.56 73.9 81.8 70.0 0.75 78.3 63.6 91.7 0.61 69.7
M5P 69.5 54.5 83.3 0.80 52.2 45.5 58.3 0.73 56.5 45.5 66.7 0.43 56.5 36.4 75.0 0.44 58.7
LDA 69.6 72.7 66.7 0.81 34.8 40.0 75.0 0.45 34.8 0.0 100 0.30 30.4 100 0.0 0.52 52.0

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