International Journal of Educational Technology in Higher Education

Table 5 Classification performance when employing DS only and combining DS and GF for datasets D1, D2, and D3

From: Extracting topological features to identify at-risk students using machine learning and graph convolutional network models

	ML Model	Accuracy					AUC
		DS	Euclidean		Cosine		DS	Euclidean		Cosine
			+GF	\(\Delta _{ACC}\)	+GF	\(\Delta _{ACC}\)		+GF	\(\Delta _{AUC}\)	+GF	\(\Delta _{AUC}\)
D1	XGB	0.780	0.840	7.692	0.800	2.564	0.915	0.932	1.930	0.933	1.959
	Light GBM	0.780	0.800	2.564	0.810	3.846	0.914	0.928	1.527	0.934	2.160
	SVM linear	0.690	0.660	− 4.348	0.650	− 5.797	0.840	0.861	2.469	0.813	− 3.273
	Extra Trees	0.790	0.820	3.797	0.770	− 2.532	0.929	0.950	2.252	0.962	3.523
	Bagging	0.630	0.670	6.349	0.670	6.349	0.838	0.853	1.769	0.830	− 1.039
	Random Forest	0.710	0.810	14.085	0.790	11.268	0.926	0.955	3.138	0.915	− 1.183
	Mean	0.730	0.767	5.023	0.748	2.616	0.894	0.913	2.181	0.898	0.358
D2	XGB	0.780	0.860	10.256	0.860	10.256	0.895	0.950	6.232	0.966	7.976
	Light GBM	0.790	0.860	8.861	0.840	6.329	0.915	0.935	2.176	0.959	4.740
	SVM linear	0.660	0.640	− 3.030	0.760	15.152	0.866	0.849	− 1.972	0.888	2.537
	Extra Trees	0.810	0.870	7.407	0.890	9.877	0.938	0.970	3.360	0.971	3.541
	Bagging	0.730	0.700	− 4.110	0.730	0.000	0.899	0.882	− 1.934	0.897	− 0.270
	Random Forest	0.780	0.840	7.692	0.870	11.538	0.927	0.963	3.828	0.972	4.822
	Mean	0.758	0.795	4.513	0.825	8.859	0.907	0.925	1.948	0.942	3.891
D3	XGB	0.900	0.920	2.222	0.950	5.556	0.951	0.979	2.967	0.978	2.800
	LightGBM	0.920	0.910	− 1.087	0.970	5.435	0.978	0.975	− 0.339	0.985	0.665
	SVM linear	0.780	0.800	2.564	0.790	1.282	0.917	0.907	− 1.052	0.917	0.036
	Extra Trees	0.960	0.920	− 4.167	0.950	− 1.042	0.995	0.980	− 1.548	0.990	− 0.530
	Bagging	0.850	0.810	− 4.706	0.860	1.176	0.940	0.945	0.531	0.944	0.477
	Random Forest	0.920	0.920	0.000	0.940	2.174	0.989	0.973	− 1.680	0.983	− 0.648
	Mean	0.888	0.880	− 0.862	0.910	2.430	0.962	0.960	− 0.187	0.966	0.467

DS: initial dataset; GF: graph topological features
\(\Delta\) shows the boost in performance achieved by adding GF to DS

Back to article page