Blood Glucose Level Prediction of Diabetic Type 1 Patients Using Nonlinear Autoregressive Neural Networks

Table 3

Comparison of feedforward and nonlinear autoregressive models.

Dataset	Source	Feedforward NN (RMSE) ml/dl	NAR NN with optimal inputs (RMSE ml/dl)
			Y(T) = 15 min	Y(t+1) = 30 min
Patient 1	AIDA	15 min = 0.9984	0.6060
		30 min = 3.78		1.127
Patient 2	AIDA	15 min = 1.43	0.7911
		30 min = 3.51		1.6756
Abalone	UCI	Y (t) = 3.36	2.417713
				2.56211
Servo	UCI	Y (t) = 1.81	0.781796
				1.493887