International Journal of Educational Technology in Higher Education

Table 6 Classification performance when employing initial D2 dataset and combining it with topological features using cosine and Euclidean metrics

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

	ML Model	Accuracy			AUC
	ML Model	DS	+GF	\(\Delta _{ACC}\), %	DS	+GF	\(\Delta _{AUC}\), %
Cosine	XGB	0.920	0.924	0.435	0.961	0.979	1.873
	Light GBM	0.890	0.924	3.820	0.967	0.983	1.655
	SVM linear	0.760	0.773	1.711	0.909	0.922	1.430
	Extra Trees	0.870	0.950	9.195	0.980	0.995	1.531
	Random Forest	0.870	0.874	0.460	0.965	0.981	1.658
	MLP	0.760	0.790	3.947	0.907	0.943	3.969
	Mean	0.845	0.873	3.261	0.948	0.967	2.019
Euclidean	XGBClassifier	0.920	0.874	− 5.000	0.961	0.963	0.208
	Light GBM	0.890	0.820	− 0.899	0.967	0.972	0.517
	SVM linear	0.760	0.798	5.000	0.909	0.928	2.090
	Extra Trees	0.870	0.882	1.379	0.980	0.975	-0.510
	Random Forest	0.870	0.891	2.414	0.965	0.983	1.865
	MLP	0.760	0.756	− 0.526	0.907	0.920	1.433
	Mean	0.845	0.847	0.395	0.948	0.957	0.934

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

Back to article page