Cognitive diagnosis method via neural networks with transfer learning and Q-matrix constraints

doi:10.3724/SP.J.1041.2026.0755

Abstract

Cite this article

TAO Jinhong, ZHAO Wei, CHENG Nuo, QIAO Lifang, JIANG Qiang. (2026). Cognitive diagnosis method via neural networks with transfer learning and Q-matrix constraints. Acta Psychologica Sinica, 58(4), 755-772.

share this article

0
/ / Recommend

Add to citation manager EndNote|Ris|BibTeX

URL: https://journal.psych.ac.cn/acps/EN/10.3724/SP.J.1041.2026.0755

https://journal.psych.ac.cn/acps/EN/Y2026/V58/I4/755

Figures/Tables 2

Table 1 Attribute Mastery Rate (AMR) of Different Models under Different Conditions

Simulated dataset	Quality	N	DINA	GDINA	NPC	GNPC	ANN	Bi-QNN
SD1	High	50	0.945	0.929	0.942	0.966	0.816	0.963
		100	0.957	0.953	0.942	0.968	0.814	0.968
		200	0.959	0.962	0.944	0.970	0.892	0.970
		300	0.962	0.964	0.944	0.972	0.929	0.972
		500	0.965	0.967	0.946	0.973	0.966	0.975
		Mean	0.958	0.955	0.944	0.970	0.883	0.970
	Low	50	0.835	0.818	0.839	0.865	0.764	0.886
		100	0.840	0.830	0.843	0.867	0.776	0.892
		200	0.85	0.844	0.844	0.876	0.810	0.904
		300	0.852	0.854	0.845	0.882	0.860	0.908
		500	0.853	0.864	0.846	0.883	0.896	0.911
		Mean	0.846	0.842	0.843	0.875	0.821	0.900
SD2	High	50	0.878	0.852	0.885	0.898	0.805	0.906
		100	0.877	0.858	0.890	0.907	0.825	0.907
		200	0.887	0.867	0.898	0.915	0.868	0.917
		300	0.889	0.880	0.899	0.920	0.886	0.921
		500	0.899	0.892	0.902	0.921	0.917	0.926
		Mean	0.886	0.870	0.895	0.912	0.860	0.915
	Low	50	0.748	0.723	0.777	0.759	0.773	0.846
		100	0.765	0.726	0.785	0.772	0.771	0.845
		200	0.792	0.748	0.785	0.776	0.815	0.855
		300	0.802	0.767	0.790	0.782	0.831	0.858
		500	0.823	0.794	0.796	0.788	0.854	0.858
		Mean	0.786	0.752	0.787	0.775	0.809	0.852
SD3	High	50	0.865	0.847	0.889	0.918	0.825	0.940
		100	0.877	0.861	0.895	0.933	0.825	0.938
		200	0.882	0.866	0.896	0.936	0.876	0.945
		300	0.888	0.873	0.899	0.938	0.897	0.947
		500	0.894	0.875	0.901	0.940	0.931	0.950
		Mean	0.881	0.864	0.896	0.933	0.871	0.944
	Low	50	0.772	0.766	0.775	0.797	0.793	0.882
		100	0.794	0.770	0.797	0.813	0.775	0.866
		200	0.811	0.771	0.803	0.817	0.825	0.882
		300	0.818	0.783	0.806	0.822	0.840	0.882
		500	0.834	0.795	0.808	0.823	0.869	0.888
		Mean	0.806	0.777	0.798	0.814	0.820	0.880

Table 1 Attribute Mastery Rate (AMR) of Different Models under Different Conditions

Simulated dataset	Quality	N	DINA	GDINA	NPC	GNPC	ANN	Bi-QNN
SD1	High	50	0.945	0.929	0.942	0.966	0.816	0.963
		100	0.957	0.953	0.942	0.968	0.814	0.968
		200	0.959	0.962	0.944	0.970	0.892	0.970
		300	0.962	0.964	0.944	0.972	0.929	0.972
		500	0.965	0.967	0.946	0.973	0.966	0.975
		Mean	0.958	0.955	0.944	0.970	0.883	0.970
	Low	50	0.835	0.818	0.839	0.865	0.764	0.886
		100	0.840	0.830	0.843	0.867	0.776	0.892
		200	0.85	0.844	0.844	0.876	0.810	0.904
		300	0.852	0.854	0.845	0.882	0.860	0.908
		500	0.853	0.864	0.846	0.883	0.896	0.911
		Mean	0.846	0.842	0.843	0.875	0.821	0.900
SD2	High	50	0.878	0.852	0.885	0.898	0.805	0.906
		100	0.877	0.858	0.890	0.907	0.825	0.907
		200	0.887	0.867	0.898	0.915	0.868	0.917
		300	0.889	0.880	0.899	0.920	0.886	0.921
		500	0.899	0.892	0.902	0.921	0.917	0.926
		Mean	0.886	0.870	0.895	0.912	0.860	0.915
	Low	50	0.748	0.723	0.777	0.759	0.773	0.846
		100	0.765	0.726	0.785	0.772	0.771	0.845
		200	0.792	0.748	0.785	0.776	0.815	0.855
		300	0.802	0.767	0.790	0.782	0.831	0.858
		500	0.823	0.794	0.796	0.788	0.854	0.858
		Mean	0.786	0.752	0.787	0.775	0.809	0.852
SD3	High	50	0.865	0.847	0.889	0.918	0.825	0.940
		100	0.877	0.861	0.895	0.933	0.825	0.938
		200	0.882	0.866	0.896	0.936	0.876	0.945
		300	0.888	0.873	0.899	0.938	0.897	0.947
		500	0.894	0.875	0.901	0.940	0.931	0.950
		Mean	0.881	0.864	0.896	0.933	0.871	0.944
	Low	50	0.772	0.766	0.775	0.797	0.793	0.882
		100	0.794	0.770	0.797	0.813	0.775	0.866
		200	0.811	0.771	0.803	0.817	0.825	0.882
		300	0.818	0.783	0.806	0.822	0.840	0.882
		500	0.834	0.795	0.808	0.823	0.869	0.888
		Mean	0.806	0.777	0.798	0.814	0.820	0.880

Table2 Attribute-Pattern Classification Consistency of Each Model on the Sub FRAC Dataset

Real Dataset	Metrics	N	DINA	GDINA	NPC	GNPC	ANN	Bi-QNN
SFRAC	AMR	50	0.920	0.906	0.840	0.854	0.887	0.923
		100	0.931	0.896	0.841	0.860	0.927	0.956
		200	0.942	0.920	0.848	0.812	0.951	0.959
		300	0.953	0.919	0.851	0.786	0.953	0.963
		500	0.953	0.929	0.848	0.773	0.950	0.960
		Mean	0.940	0.914	0.846	0.817	0.934	0.952
	PMR	50	0.706	0.664	0.508	0.500	0.572	0.674
		100	0.742	0.630	0.530	0.541	0.701	0.803
		200	0.761	0.710	0.533	0.448	0.777	0.820
		300	0.807	0.700	0.550	0.421	0.795	0.836
		500	0.799	0.736	0.548	0.392	0.785	0.829
		Mean	0.763	0.688	0.534	0.460	0.726	0.792
	PMR(K-1)	50	0.908	0.872	0.756	0.798	0.890	0.946
		100	0.926	0.874	0.754	0.800	0.943	0.979
		200	0.951	0.905	0.768	0.709	0.978	0.976
		300	0.960	0.903	0.779	0.661	0.973	0.980
		500	0.966	0.917	0.772	0.613	0.964	0.974
		Mean	0.942	0.894	0.766	0.716	0.950	0.971

Table2 Attribute-Pattern Classification Consistency of Each Model on the Sub FRAC Dataset

Real Dataset	Metrics	N	DINA	GDINA	NPC	GNPC	ANN	Bi-QNN
SFRAC	AMR	50	0.920	0.906	0.840	0.854	0.887	0.923
		100	0.931	0.896	0.841	0.860	0.927	0.956
		200	0.942	0.920	0.848	0.812	0.951	0.959
		300	0.953	0.919	0.851	0.786	0.953	0.963
		500	0.953	0.929	0.848	0.773	0.950	0.960
		Mean	0.940	0.914	0.846	0.817	0.934	0.952
	PMR	50	0.706	0.664	0.508	0.500	0.572	0.674
		100	0.742	0.630	0.530	0.541	0.701	0.803
		200	0.761	0.710	0.533	0.448	0.777	0.820
		300	0.807	0.700	0.550	0.421	0.795	0.836
		500	0.799	0.736	0.548	0.392	0.785	0.829
		Mean	0.763	0.688	0.534	0.460	0.726	0.792
	PMR(K-1)	50	0.908	0.872	0.756	0.798	0.890	0.946
		100	0.926	0.874	0.754	0.800	0.943	0.979
		200	0.951	0.905	0.768	0.709	0.978	0.976
		300	0.960	0.903	0.779	0.661	0.973	0.980
		500	0.966	0.917	0.772	0.613	0.964	0.974
		Mean	0.942	0.894	0.766	0.716	0.950	0.971

References

Metrics

Viewed

Full text

Abstract

Comments

Abstract
Figures/Tables
References
Related Articles
Recommended Articles
Metrics
Comments

TOP