Identification of Insider Trading in the Securities Market Based on Multi-task Deep Neural Network

Table 6

Comparison of multiple models in insider trading identification.

Methods		Logistic		SVM		RF		XGBoost		DNN
		Pool	STL	Pool	STL	Pool	STL	Pool	STL	Pool	STL	MTL
Accuracy	Industry1	—	0.837	—	0.811	—	0.963	—	0.974	—	0.889	0.868
	Industry2	—	0.849	—	0.939	—	0.849	—	0.859	—	0.846	0.969
	Industry3	—	0.875	—	0.854	—	0.965	—	0.931	—	0.667	0.909
	Industry4	—	0.794	—	0.830	—	0.766	—	0.765	—	0.788	0.960
	Industry5	—	0.800	—	0.789	—	0.903	—	0.974	—	0.902	0.966
	Average	0.856	0.831	0.813	0.845	0.901	0.889	0.874	0.900	0.901	0.818	0.934
Recall	Industry1	—	0.615	—	0.732	—	0.867	—	0.946	—	0.889	0.871
	Industry2	—	0.529	—	0.861	—	0.652	—	0.582	—	0.861	0.911
	Industry3	—	0.750	—	0.940	—	0.843	—	0.615	—	0.708	0.893
	Industry4	—	0.511	—	0.760	—	0.538	—	0.500	—	0.789	0.939
	Industry5	—	0.600	—	0.952	—	0.819	—	0.964	—	0.903	0.917
	Average	0.682	0.601	0.858	0.849	0.758	0.744	0.724	0.721	0.841	0.830	0.906
AUC	Industry1	—	0.782	—	0.771	—	0.914	—	0.973	—	0.892	0.871
	Industry2	—	0.868	—	0.890	—	0.734	—	0.869	—	0.868	0.967
	Industry3	—	0.920	—	0.750	—	0.909	—	0.977	—	0.685	0.896
	Industry4	—	0.769	—	0.915	—	0.651	—	0.690	—	0.790	0.960
	Industry5	—	0.726	—	0.722	—	0.911	—	0.987	—	0.902	0.966
	Average	0.763	0.813	0.791	0.810	0.879	0.824	0.902	0.899	0.907	0.828	0.932
F1-score	Industry1	—	0.667	—	0.826	—	0.836	—	0.972	—	0.930	0.878
	Industry2	—	0.546	—	0.872	—	0.634	—	0.650	—	0.949	0.969
	Industry3	—	0.692	—	0.836	—	0.906	—	0.762	—	0.874	0.973
	Industry4	—	0.632	—	0.867	—	0.557	—	0.533	—	0.763	0.958
	Industry5	—	0.643	—	0.936	—	0.825	—	0.947	—	0.921	0.945
	Average	0.677	0.636	0.814	0.867	0.789	0.752	0.846	0.773	0.898	0.887	0.944