A Novel Clustering-Based Algorithm for Continuous and Noninvasive Cuff-Less Blood Pressure Estimation

Table 2

Comparison of the performance with clustering and regression algorithms.

			Systolic blood pressure (mmHg)			Diastolic blood pressure (mmHg)
Learner/performance	Cluster	Count of data per cluster	MAE	RMSE	r	MAE	RMSE	r
Random forest regression	Cluster1	6282	3.407	5.830	—	3.250	5.698	—
	Cluster2	6276	3.468	5.724	—	3.038	5.136	—
	Cluster3	3355	3.521	5.586	—	2.813	5.408	—
	Cluster4	8300	3.396	5.500	—	2.870	4.852	—
	Cluster5	2390	2.434	4.567	—	2.677	4.879	—
	Total	26,603	3.344	5.557	—	2.974	5.191	—
Gradient boosting regression	Cluster1	6282	2.644	5.841	0.96	2.486	5.648	0.98
	Cluster2	6276	2.781	5.694	0.93	2.468	5.232	0.96
	Cluster3	3355	2.533	6.123	0.76	2.003	5.491	0.80
	Cluster4	8300	2.610	5.522	0.85	2.161	4.675	0.95
	Cluster5	2390	1.643	4.709	0.85	1.504	4.467	0.95
	Total	26,603	2.561	5.635	0.88	2.231	5.012	0.94
Multilayer perceptron regression	Cluster1	6282	5.230	8.244	—	4.896	7.262	—
	Cluster2	6276	5.340	8.754	—	5.263	8.956	—
	Cluster3	3355	6.235	9.523	—	6.094	8.852	—
	Cluster4	8300	7.261	11.920	—	6.288	9.003	—
	Cluster5	2390	4.326	8.156	—	4.160	6.875	—
	Total	26,603	5.937	9.664	—	5.501	8.370	—