Development and Internal Validation of Machine Learning Algorithms for Determining Sodium Valproate Concentrations below the Standard Level Using a Risk Prediction Model of Children with Epilepsy

Table 1

Forecast results for development group and validation group of machine learning algorithms.

	Sen	Spe	TP	FN	FP	TN	Accuracy	Precision	Recall	F1 score
Development group (standard/low = 297/124)
RF	1.00	1.00	124	0	0	297	1.00	1.00	1.00	1.00
SVM	0.806	0.771	96	28	58	239	0.796	0.623	0.774	0.691
GLM	0.718	0.636	79	45	84	213	0.694	0.485	0.637	0.551
GBM	0.734	0.815	101	23	79	218	0.758	0.561	0.815	0.664
NNET	1.00	0.24	3	121	0	297	0.713	1.000	0.024	0.047
Validation group (standard/low = 80/24)
RF	1.00	1.00	24	0	0	80	1.00	1.00	1.00	1.00
SVM	0.958	0.912	23	1	7	73	0.923	0.767	0.958	0.852
GLM	0.667	0.662	16	8	27	53	0.663	0.372	0.667	0.478
GBM	0.833	0.737	20	4	21	59	0.760	0.488	0.833	0.615
NNET	0.625	0.612	15	9	31	49	0.615	0.326	0.625	0.429

RF: random forest; SVM: support vector machine; GBM: gradient boosting machine; GLM: generalized linear model; NNET: neural network; Sen: sensitivity; Spe: specificity; TP: true positive; FN: false negative; FP: false positive; TN: true negative.