Predicting Student Academic Performance at Higher Education Using Data Mining: A Systematic Review

Table 2

Summary of the related studies on predicting student’s level in the course.

Ref	Techniques	Results	Study sample size	Findings
[29]	RF & ANN	Accuracy (74%)	1854	The results showed that the RF and ANN outperformed SVM, LR, NB, and KNN.
[31]	RF	Accuracy (79%–91%)	32,593	The feature engineering technique improved the prediction model where it achieved more than 80% accuracy.
[32]	Ensemble (bagging)	Accuracy (66.7%) for dataset_1	Dataset_1 : 52	The highest results were achieved when considering 50% of the coursework.
		(93.1%) for dataset_2	Dataset_2 : 486
[33]	LR	Accuracy (78.75%)	499	Several factors including data cleaning, the type of features, and the dataset size influenced the final model’s accuracy.
[34]	RF	Accuracy (59.64%)	289	The formative assessment tasks grades were able to predict at-risk students.
[35]	Hybrid (DT and K-means)	Accuracy (75.47%)	—	Student’s behavior and students’ academic success are very strongly related.
[36]	RF	Accuracy (96.4%)	145	Regression methods outperformed the classification in predicting course GPA.
[37]	BP-NN	—	101	NCEE score and learning attitude have a major influence on the final exam results.
[38]	ANN (MLP)	Accuracy (76.07%)	163	—
[30]	Ensemble (bagging)	Accuracy (99.40%)	480	Ensemble models outperformed other classification algorithms.
[39]	ANN	Accuracy (78.1%)	480	Student’s attendance and parents’ participation have a great impact on student’s performance level.
[40]	SVM	Accuracy (76.3%)	38	Preadmission information influences the student grades
[41]	Ensemble method with DT	Accuracy (95.78%)	1044	A balanced dataset and the selected features increase the model’s accuracy.
[42]	RF	F1-score (74%)	240	The accuracy of the prediction model has improved when considering the 2^nd test.
[43]	RF	MAE (1.198 to 1.91)	592	RF provides the best result followed by SMOreg and bagging.
[44]	SVM	F1-score (92%) for distance dataset	Online dataset: 262	The accuracy of the prediction model has improved when the student performed 50% of distance education course and 25% of on-campus education course.
		F1-score (83%) for on-campus dataset	On-campus dataset: 161
[45]	RBM	RMSE (0.3)	225	CGPA, HSSC, and an entry test can predict student performance.
[46]	NB	—	700	Demographic information and school grades allow predicting a college student’s level.