Applied Network Science

Table 1 Entity classification accuracy for baseline models and different pruning methods on random walks on three classification models: k-nearest neighbors (k-NN), support vector machine (SVM), and random forest (RF)

From: Fingerprinting Bitcoin entities using money flow representation learning

	Mining Pools			BitcoinHeist			WalletExplorer
	k-NN	SVM	RF	k-NN	SVM	RF	k-NN	SVM	RF
Baseline (Freq. 1% Clusters)
Graph2Vec	0.1480	0.2693	0.2200	0.3497	0.4124	0.4813	0.2494	0.3779	0.3264
SVD	0.6040	0.6253	0.7613	0.4977	0.3191	0.5846	0.3924	0.3536	0.3933
Random Walk
Freq. 1% clusters	0.6467	0.9613	0.8547	0.5627	0.6921	0.5766	0.1900	0.5948	0.3958
Freq. 5% clusters	0.7520	0.9493	0.8373	0.5941	0.7023	0.5766	0.2252	0.6282	0.4121
Freq. 10% clusters	0.7800	0.9493	0.8147	0.4555	0.6992	0.5773	0.2236	0.6148	0.3985
Freq. 20% clusters	0.7987	0.9467	0.8267	0.5420	0.6676	0.5752	0.2373	0.5967	0.3955
Known entity name	0.6947	0.8187	0.6720	0.4210	0.5178	0.5126	0.3591	0.5370	0.4006
Known entity type	0.6360	0.6893	0.6933	0.3173	0.3664	0.3926	0.4606	0.5285	0.4433

Bold highlights the highest score overall, and itatic highlights the highest baseline score for each dataset

Back to article page