Table 4 Results of detection with different embedding methods
From: Aparecium: understanding and detecting scam behaviors on Ethereum via biased random walk
Method Category | Method | Precision | Recall | F1-score | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Versatile Method | Deepwalk (Jin et al. 2022) | 0.911 | 0.729 | 0.810 | |||||||||
Node2vec (Zhou et al. 2021) | 0.864 | 0.842 | 0.853 | ||||||||||
GraphSage (Huang et al. 2022) | 0.949 | 0.851 | 0.897 | ||||||||||
GCN (Patel et al. 2020a) | 0.984 | 0.848 | 0.911 | ||||||||||
For Blockchain | Trans2vec (Wu et al. 2020) | 0.905 | 0.823 | 0.862 | |||||||||
T-EDGE (Lin et al. 2020b) | 0.878 | 0.776 | 0.824 | ||||||||||
I\(^2\)BGNN (Shen et al. 2021) | 0.869 | 0.903 | 0.886 | ||||||||||
MCGC (Zhang et al. 2021) | 0.874 | 0.901 | 0.887 | ||||||||||
Our Method | Network Structure Bias | 0.887 | 0.776 | 0.830 | |||||||||
Pure Semantic Bias | 0.877 | 0.792 | 0.832 | ||||||||||
Time Mixed Bias | 0.869 | 0.808 | 0.837 | ||||||||||
All | 0.977 | 0.957 | 0.967 | ||||||||||