Skip to main content

Table 5 MATADOR random-slice results

From: Neural networks for link prediction in realistic biomedical graphs: a multi-dimensional evaluation of graph embedding-based approaches

  Node AUC AUC   Avg. Prec
Method combination (ROC) (PR) MAP R-prec @ k
Deep- Average 95.93 95.82 89.81 86.86 98.77*
Walk Concat 94.97 94.83 88.30 84.63 98.34*
LINE Average 80.63 81.30 67.74 61.04 91.65
  Concat 81.16 81.82 68.53 61.42 92.00
node- Average 78.38 78.75 66.42 59.32 88.67
2vec Concat 77.62 77.54 65.44 58.40 87.25
AA N/A 91.97 88.40 87.16 85.06 86.87
CN N/A 97.27 97.04* 95.47 94.64 98.74*
JI N/A 97.23* 97.10 94.72 92.29 98.96
  1. (Bold: best score, *: not statistically different from best)