Table 5 Gap between the best AUC value and the AUC of the algorithm corresponding to the best measure value
From: Unsupervised evaluation of multiple node ranks by reconstructing local structures
| Â | Â | Structural Measures | Local Structure Measures | Â | ||||
|---|---|---|---|---|---|---|---|---|
Network | Examples | Cond/nce | Density | Modularity | LinkCC | LinkAUC | HopAUC | NDCG |
Amazon | 0.1% | 24% | 23% | 23% | 0% | 0% | 0% | 0% |
Amazon | 1% | 20% | 6% | 12% | 6% | 0% | 0% | 3% |
Amazon | 10% | 9% | 4% | 6% | 7% | 1% | 2% | 0% |
Amazon | 25% | 5% | 4% | 27% | 0% | 2% | 2% | 1% |
Amazon | 50% | 17% | 3% | 3% | 3% | 2% | 2% | 1% |
DBLP | 0.1% | 11% | 10% | 6% | 9% | 6% | 6% | 0% |
DBLP | 1% | 4% | 11% | 1% | 15% | 0% | 0% | 0% |
DBLP | 10% | 1% | 5% | 1% | 2% | 1% | 1% | 5% |
DBLP | 25% | 0% | 2% | 0% | 2% | 0% | 1% | 2% |
DBLP | 50% | 1% | 2% | 1% | 1% | 1% | 1% | 0% |
BlockModel | 0.1% | 33% | 12% | 33% | 33% | 5% | 5% | 0% |
BlockModel | 1% | 29% | 7% | 0% | 36% | 6% | 6% | 0% |
BlockModel | 10% | 33% | 8% | 4% | 40% | 13% | 5% | 0% |
BlockModel | 25% | 35% | 5% | 9% | 43% | 10% | 10% | 0% |
BlockModel | 50% | 36% | 4% | 1% | 42% | 9% | 1% | 1% |
CiteSeer | 0.1% | 4% | 1% | 0% | 1% | 1% | 1% | 1% |
CiteSeer | 1% | 1% | 0% | 1% | 4% | 3% | 4% | 2% |
CiteSeer | 10% | 2% | 2% | 2% | 1% | 2% | 2% | 0% |
CiteSeer | 25% | 3% | 2% | 3% | 3% | 3% | 3% | 1% |
CiteSeer | 50% | 3% | 4% | 3% | 3% | 3% | 3% | 1% |
PubMed | 0.1% | 9% | 13% | 2% | 22% | 19% | 18% | 0% |
PubMed | 1% | 7% | 9% | 7% | 21% | 12% | 7% | 1% |
PubMed | 10% | 4% | 6% | 4% | 4% | 9% | 9% | 0% |
PubMed | 25% | 4% | 8% | 4% | 27% | 4% | 4% | 0% |
PubMed | 50% | 5% | 11% | 5% | 27% | 4% | 3% | 0% |
Number of gaps ≤5% | 11 | 13 | 18 | 12 | 17 | 20 | 25 | |
Number of gaps ≤7% | 12 | 16 | 22 | 14 | 19 | 23 | 25 | |