Extracting matching features using tMatchModel
You can use the labeled sample of suspect duplicate pairs as the input of the
tMatchModel component.
You have to specify the set of columns the model will be built on and the column specifying the label. The algorithm will compute different measures, called features, to catch as much information as possible on this set of columns.
The tMatchModel component uses the Random forest algorithm to build the model. This algorithm is a generalization of decision trees.