Table 1.

Performance of the C4.5 decision tree, logistic regression, and support vector machine models on the training set and test set.

Model	Data set	P/N	$n$	ACC	SE	SP	AUC	MCC
C4.5 Decision tree	Training set	43/133	176	0.95	0.90	0.96	0.95	0.86
	Test set	20/44	64	0.92	0.83	0.98	0.95	0.83
Logistic regression	Training set	43/133	176	0.86	0.73	0.90	0.77	0.56
	Test set	20/44	64	0.80	0.73	0.82	0.77	0.58
Support vector machine (Polynomial kernel)	Training set	43/133	176	0.85	0.74	0.87	0.76	0.61
	Test set	20/44	64	0.83	0.80	0.84	0.77	0.50

Note: Predictive accuracy was reflected by four indices: sensitivity ($\text{SE}=\text{TP}/\left[\text{TP\hspace{0.17em}}+\text{\hspace{0.17em}FN}\right]$), specificity ($\text{SP}=\text{TN}/\left[\text{TN\hspace{0.17em}}+\text{\hspace{0.17em}FP}\right]$), overall predictive accuracy ($\text{ACC}=\left[\text{TP\hspace{0.17em}}+\text{TN}\right]/\left[\text{TP\hspace{0.17em}}+\text{\hspace{0.17em}FP\hspace{0.17em}}+\text{\hspace{0.17em}TN\hspace{0.17em}}+\text{\hspace{0.17em}FN}\right]$), and Matthews correlation coefficient $(\text{MCC}=\frac{\text{TP}\times \text{TN}-\text{\hspace{0.17em}FP}\times \text{FN}}{\sqrt{(\text{TP\hspace{0.17em}}+\text{\hspace{0.17em}FP})(\text{TP\hspace{0.17em}}+\text{\hspace{0.17em}FN})(\text{TN\hspace{0.17em}}+\text{\hspace{0.17em}FP})(\text{TN\hspace{0.17em}}+\text{\hspace{0.17em}FN})}})$. The area under the receiver operating characteristic curve (AUC) is a measure of how well a model distinguishes positive and negative data points; the $>95\%$ model AUC illustrated a high classification power. FN, false negatives; FP, false positives; $n$, number of data points in the data set; P/N, ratio of positive/negative data points; TN, true negatives; TP, true positives.