GFD: A Weighted Heterogeneous Graph Embedding Based Approach for Fraud Detection in Mobile Advertising

Table 3

Classification results for all embedding models in AUC (mean ± std).

Date	DeepWalk	Node2vec	Metapath2vec	WMP2vec
1st June	0.788 ± 0.027	0.801 ± 0.038	0.837 ± 0.038	0.877 ± 0.030
2nd June	0.828 ± 0.032	0.848 ± 0.025	0.864 ± 0.028	0.893 ± 0.027
3rd June	0.935 ± 0.023	0.950 ± 0.024	0.912 ± 0.028	0.921 ± 0.024
4th June	0.824 ± 0.032	0.824 ± 0.040	0.849 ± 0.045	0.879 ± 0.045
5th June	0.799 ± 0.060	0.835 ± 0.055	0.853 ± 0.052	0.849 ± 0.038
6th June	0.757 ± 0.076	0.891 ± 0.042	0.844 ± 0.046	0.856 ± 0.041
7th June	0.804 ± 0.031	0.820 ± 0.027	0.844 ± 0.038	0.825 ± 0.049