Precision education with statistical learning and deep learning: a case study in Taiwan

Table 3 Accuracy, sensitivity, and specificity of the two models with different thresholds

Threshold	Prediction model	Accuracy^a	Sensitivity^b	Specificity^c
10%	Multilayer perceptron	77%	53%	80%
10%	Logistic regression	68%	61%	69%
15%	Multilayer perceptron	83%	42%	88%
15%	Logistic regression	80%	50%	84%
20%	Multilayer perceptron	87%	42%	93%
20%	Logistic regression	84%	38%	90%
50%	Multilayer perceptron	90%	16%	99%
50%	Logistic regression	88%	0%	100%

^aAccuracy = predicted number (dropout students + current students) / actual number (dropout students + current students)
^bSensitivity = predicted number of dropout students / actual number of dropout students
^cSpecificity = predicted number of current students / actual number of current students