Table 1 Comparison of Methodologies and Results from Various Studies.

S1⁶

Optimised SMOTE+ (ML) classifiers.

SDP dataset

0.99

0.68 (NN)

1.00 (NN)

0.91 (NN)

0.98 (NN)

S2⁷

SVM-RBF

Czech Technical University Dataset

0.77 ± 0.05

0.69 ± 0.08

0.64 ± 0.09

0.90 ± 0.09

S3⁸

DeepFM

HarvardX Person-Course Academic Year 2013 De-Identified Dataset.

99%

0.98

0.99

0.98

0.92

S4⁹

SMOTE + RF, SVMSMOTE + RF

Data collected from internal and external sources

0.741 ± 0.005

0.84 ± 0.02

S5¹⁰

Multi-feature fusion algorithm

KDD CUP 2015 Dataset

0.8885

0.899

0.9627

0.9311

S6¹¹

LR, DT, RF, SVM, DNN, LightGBM

2013-2022 with 20,050 student records.

0.955 (LightGBM)

0.867 (LightGBM)

0.814 (LightGBM)

0.840 (LightGBM)

S7¹²

LR, DT, RF, LightGBM, SVM, XGBoost

School Dataset

0.94

0.9523

0.9393

0.9703

0.895

S8¹³

AutoML. Grid search and randomized search

Twaweza Uwezo Dataset (168,162 samples, 15 features)

0.99 (RF, grid search)

0.84

0.85

0.82

S9¹⁴

CNN, LSTM, ensemble approaches

KDD CUP 2015 dataset

0.91896 (Bagging LSTM-LSTM)

0.919 (Bagging LSTM-LSTM)

0.9899 (Bagging LSTM-LSTM)

0.98 (Bagging LSTM-LSTM)