FIG. 3.

Median points of TPR (in square), precision (in triangle), and FPR (in cross) under large feature space data for three BN leaners: bAIcis, bnlearn, and sparsebn. It is an integrated figure composed of eight subfigures stratified by network type and topology structure (continuous random topology network, continuous scale-free topology network, discrete random topology network, and mixed random topology network) vertically and node size (500 and 2000) horizontally; in each subfigure, y-axis presents the metric value ranging from 0 to 1, x-axis shows the compared BN tools, and median points are further stratified by sample size (50 in brown, 200 in green and 1000 in yellow). The variance among the replicates was extremely small and hence was not shown. The outcomes for bnlearn on the continuous networks with 2000 nodes are left blank due to its unexpected long running time.