Table 2 Performances comparison with different negative samples selecting strategies (random strategy is denoted “random”, our strategy is “reliable”)

meta-path-1

0.894

0.896

0.859

0.861

0.786

0.771

0.875

0.891

0.835

0.835

0.673

0.677

0.827

0.826

meta-path-2

0.930

0.936

0.925

0.928

0.873

0.850

0.839

0.873

0.850

0.861

0.702

0.722

0.855

0.860

meta-path-3

0.921

0.926

0.902

0.905

0.826

0.832

0.862

0.883

0.843

0.858

0.690

0.719

0.842

0.855

meta-path-4

0.894

0.895

0.858

0.860

0.782

0.790

0.882

0.867

0.836

0.832

0.676

0.667

0.828

0.826

meta-path-5

0.918

0.920

0.892

0.895

0.809

0.800

0.900

0.925

0.859

0.865

0.721

0.737

0.852

0.858

ensemble

0.954

0.956

0.949

0.951

0.924

0.913

0.837

0.854

0.871

0.876

0.745

0.755

0.878

0.882