A Novel CGM Metric-Gradient and Combining Mean Sensor Glucose Enable to Improve the Prediction of Nocturnal Hypoglycemic Events in Patients with Diabetes

Table 2

The influence of CGM length on the specificity and sensitivity of different algorithms when the metric of MSG+gradient was adopted.

pH	Algorithm	CGM		CGM		CGM
		SPE (%)	SEN (%)	SPE (%)	SEN (%)	SPE (%)	SEN (%)
15 min	LR	97.12	98.01	97.47	97.70	96.85	97.35
	SVM	97.40	98.01	97.41	97.35	96.26	97.35
	RF	96.07	96.03	95.92	95.36	95.30	95.36
	LSTM	96.88	97.35	97.10	97.35	96.46	96.69
30 min	LR	90.83	90.07	91.18	90.07	90.73	91.39
	SVM	90.30	90.73	90.08	90.07	90.09	90.73
	RF	87.79	90.07	87.53	90.73	86.95	90.07
	LSTM	90.20	90.07	90.87	90.73	89.30	90.73

PH: prediction horizon; SPE: specificity; SEN: sensitivity; LR: logistic regression; SVM: support vector machine; RF: random forest; LSTM: long short-term memory.