|
| Ref | Techniques | Results | Study sample size | Findings |
|
| [52] | SVM for model 1 and model 2 | Model 1: AUC (77%) | 9652 | They found that SVM outperformed RF, DT, and ANN. |
| RF and SVM for model 3 | Model 2: AUC (91%) |
| Model 3: AUC (93%) |
| [53] | DT (CART) | Accuracy (80%) | 2407 | Mother’s job, department, father’s job, the main source of living expenseS, and the admission status are the highest influential features. |
| [54] | Ensemble | Accuracy (75.9%) | 1491 | The accuracy of the prediction model was improved after applying the SMOTE technique to balance the proposed dataset. |
| [17] | Ensemble | Accuracy (97%) | 233 | The proposed ensemble model outperformed bagging, stacking, NB, SVM, and DT. |
| [15] | ANN and DT | ANN model: Accuracy (79%) | 1,569 | The SAAT is the most influence factor on the CGPA as it scored the highest correlation coefficient among the admission attributes. |
| Precision (81%). |
| DT model: Recall (80%) F1-Measure (81%). |
| [55] | RF | F1-score, precision and recall (97.1%) | 9,458 | The SMO outperformed other classifiers before eliminating the misclassified samples while the RF scored the highest performance after eliminating the misclassified observations. |
| [56] | ANN | Accuracy (51.9%) | 1445 | The relation between admission criteria and student academic success is weak. |
| [57] | Structural equation modeling (SEM) | — | 519 | There is no relation between the noncognitive attributes and student performance. |
| [58] | MLP-ANN | Accuracy (85%) | 300 | First year GPA is the most influencing attribute on student’s performance in the second year. |
| [16] | DT (J48) | Precision (62.9%), | 161 | Student demographic features are not highly correlated with the class attribute, whereas GPA, credits, and father’s work have a significant impact on the target attribute. |
| Recall (63.4%) |
| [59] | Improved ID3 | Accuracy (74%) | 50 | — |
| [60] | SVM | Accuracy (90.60%) | 309 | The results show that the performance of machine learning and deep learning techniques are similar. |
|
