A Comparison of Regression Tree Approaches to Modelling the Efficacy of Water Hyacinth Biocontrol Using Multitemporal Spectral Datasets

Table 3

Predictive performance of the MTRTs and RF-MTRTs for low, medium, and high infestation levels.

		Multitarget regression trees		RF multitarget regression trees
		Correlation coefficient (%)	Root mean square error (FSD—number of feeding scars; RLCC-SPAD units)	Correlation coefficient (%)	Root mean square error (FSD—number of feeding scars; RLCC-SPAD units)
Infestation level	Target	Train/validation	Train/validation	Train/validation	Train/validation
Low	FSD	74/45	0.28/0.41	95/55	0.16/0.36
	RLCC	80/57	4.49/6.36	96/63	2.56/5.87
Medium	FSD	87/46	0.17/0.35	96/63	0.12/0.28
	RLCC	88/59	3.63/6.73	97/78	2.19/4.99
High	FSD	85/50	25.34/44.66	98/70	15.23/35.55
	RLCC	90/77	5.81/9.25	97/88	2.83/6.33

RLCC: relative leaf chlorophyll content; FSD: feeding scar damage.