Table 3. Predictive Capability for the Data-Fusion Strategies^a.

		training data set
literature ID as test data	cluster ID	same reaction mechanismb	allc	in the clusterd	outside the clustere
028	0	0.460	0.362	0.457	0.385
129	1	0.573	0.440	-f	(0.440)
230	0	0.432	0.400	0.411	0.445
331	2	0.842	0.378	0.312	0.792
432	2	0.949	0.500	0.171	0.925
533	2	-f	0.776	0.731	0.700
634	2	0.512	0.505	0.539	0.857

^aThe MAE values [kcal mol^–1] for the literature data set by data-fusion strategies are shown. The best performance is highlighted in bold.

^bOnly reactions classified under the same putative reaction mechanism are shown in Table 1.

^cAll of the reactions in the literature data set.

^dReactions in the same cluster in the ECFP6-based reaction descriptor space.

^eReactions not belonging to the cluster where the test reactions belonged.

^fNo training data.