Integration Strategy Is a Key Step in Network-Based Analysis and Dramatically Affects Network Topological Properties and Inferring Outcomes

Table 1

Performance of the classifiers constructed by seven machine learning integration strategies.

Strategy	ACC	AUC	Precision	Recall	FP rate	TP/FP
Naive Bayes	0.5391	0.62	0.524	0.539	0.518	1.041
Bayesian Networks	0.6325	0.736	0.683	0.632	0.418	1.512
Logistic Regression	0.7188	0.772	0.724	0.719	0.275	2.615
SVM	0.7144	0.723	0.738	0.714	0.267	2.674
Random Tree	0.6568	0.648	0.656	0.657	0.35	1.877
Random Forest	0.7196	0.787	0.72	0.72	0.292	2.466
J48	0.6808	0.671	0.681	0.681	0.323	2.108

Note: ACC stands for the accuracy of the correctly classified items (after a 10-fold cross-validation). AUC indicates the area under the ROC curve. Precision is the number of true positives divided by the total number of elements labelled as belonging to the positive class. Recall (also referred to as the True Positive Rate) represents the number of true positives divided by the total number of elements that actually belong to the positive class. The FP rate indicates the false positive rate. TP/FP reveals the true positive to the false positive ratio. Bold type indicates the maximum value in the ACC, AUC, Precision, Recall, and TP/FP columns and indicates the minimum value in the FP rate column.