Table 3.

Performance statistics of single and multi-model ensemble predictions with respect to the baseline (training) data and post-intervention (validation) data.

Site	Model	Weighta	Diversity	AE	RE	ReRMSE	Improvement (MM > SM)
Baseline performance
Malindi	EPIFIL	0.60	2.34	3.98	0.16	0.45	−0.07
	LYMFASIM	0.29	7.46	5.58	0.22	0.60	0.21
	TRANSFIL	0.11	7.03	8.97	0.36	1.43	0.67
	MM Ensemble	–	3.08	4.15	0.17	0.48	–
Nanaha	EPIFIL	0.32	5.25	11.69	0.19	0.81	−0.28
	LYMFASIM	0.41	10.10	9.71	0.16	0.90	−0.15
	TRANSFIL	0.27	8.23	11.70	0.19	1.43	0.28
	MM Ensemble	–	8.91	10.83	0.18	1.04	–
Pondicherry	EPIFIL	0.17	1.14	2.26	0.27	0.88	0.22
	LYMFASIM	0.75	1.86	1.20	0.14	0.56	−0.23
	TRANSFIL	0.08	5.14	3.61	0.43	0.91	0.25
	MM Ensemble	–	2.41	1.61	0.19	0.68	–
Post-intervention performance
Malindi	EPIFIL	–	0.50	3.78	0.41	0.90	0.01
	LYMFASIM	–	2.47	4.81	0.52	0.83	−0.07
	TRANSFIL	–	3.93	3.25	0.35	0.71	−0.26
	MM Ensemble	–	2.10	3.87	0.42	0.89	–
Nanaha	EPIFIL	–	1.83	8.93	0.49	1.33	0.20
	LYMFASIM	–	7.01	23.23	1.28	1.10	0.03
	TRANSFIL	–	8.53	8.04	0.44	0.72	−0.48
	MM Ensemble	–	13.64	12.24	0.68	1.07	–
Pondicherry	EPIFIL	–	1.56	2.93	0.53	1.13	0.18
	LYMFASIM	–	0.96	0.70	0.13	0.90	−0.02
	TRANSFIL	–	2.70	1.97	0.36	0.95	0.03
	MM Ensemble	–	1.89	1.66	0.30	0.93	–

^aWeights derived during calibration are used in validation.