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

Figure 2

ROC curves for seven machine learning integration strategies using 10-fold cross-validation against the gold standard datasets. Each point on the ROC curves of the seven integration strategies is created by the unique sensitivity and specificity against a specific likelihood ratio cut-off. Each name of the curve derived from the different integration strategies is shown in the legend. The different colours stand for the different curves for the different strategies. The area under the curve is also presented in the figure. Sensitivity and specificity are calculated during the 10-fold cross-validations.