Table 5

Results of PCA analgesic consumption (PCA only) prediction
PCA Analgesic Consum. Prediction (%) C4.5 bagging C4.5 AdaBoost C4.5 ANN* Random Forest Rotation Forest SVM NB
Low Consum. Sensitivity 84.3 79.0 76.9 89.2 95.4 84.1 99.9 81.4
Med Consum. Sensitivity 65.8 54.3 51.5 19.1 47.2 60.6 0.0 48.1
High Consum. Sensitivity 47.5 45.1 45.4 8.4 31.8 51.0 0.0 50.8
Low Consum. Precision 81.7 77.4 76.1 62.0 73.2 80.5 52.8 75.4
Med Consum. Precision 60.7 53.6 51.2 20.7 61.4 59.7 0.0 55.6
High Consum. Precision 75.1 52.3 49.3 22.8 85.7 68.2 0.0 51.2
Overall Accuracy 73.1 66.1 64.1 54.7 70.6 71.7 52.7 65.4

*ANN consisting of an input layer of 279 input units, one hidden layer of 140 hidden units, and one output layer of 3 output units.

Learning rate= 0.3; momentum rate= 0.2.

SVM using a radial basis function, exp(−gamma*|u-v|2), where gamma=1/(number of attributes)=1/279.

Hu et al.

Hu et al. BMC Medical Informatics and Decision Making 2012 12:131   doi:10.1186/1472-6947-12-131

Open Data