Table 1 The summary of machine learning... | Emerald Publishing

Table 1

The summary of machine learning algorithms for predicting cheating from exam results

Approach	Algorithms/models	Data used	Key strengths	Performance highlights
Exam score anomaly detection	RNN + Outlier Detection	Sequential exam scores	Captures temporal patterns, unsupervised	TPR∼95%, FPR∼5%
Binary classification	GLM, LR, DT, RF	Demographics + psychological features	Interpretable, uses fraud theory factors	Precision 50–75% for cheating class
Deep learning temporal models	DNN, LSTM, DenseLSTM, RNN	Exam score sequences	High accuracy on complex patterns	Up to 95% accuracy
Exam score anomaly detection	RNN + Outlier Detection	Sequential exam scores	Captures temporal patterns, unsupervised	TPR∼95%, FPR∼5%
Video behavior detection	Improved YOLOv8 + Attention mechanism	Exam room video footage	Real-time detection of cheating actions	∼82.7% accuracy

Approach	Algorithms/models	Data used	Key strengths	Performance highlights
Exam score anomaly detection	RNN + Outlier Detection	Sequential exam scores	Captures temporal patterns, unsupervised	TPR∼95%, FPR∼5%
Binary classification	GLM, LR, DT, RF	Demographics + psychological features	Interpretable, uses fraud theory factors	Precision 50–75% for cheating class
Deep learning temporal models	DNN, LSTM, DenseLSTM, RNN	Exam score sequences	High accuracy on complex patterns	Up to 95% accuracy
Exam score anomaly detection	RNN + Outlier Detection	Sequential exam scores	Captures temporal patterns, unsupervised	TPR∼95%, FPR∼5%
Video behavior detection	Improved YOLOv8 + Attention mechanism	Exam room video footage	Real-time detection of cheating actions	∼82.7% accuracy