A common form of prior knowledge in economic modelling concerns the monotonicity of relations between the dependent and explanatory variables. Monotonicity may also be an important requirement with a view toward explaining and justifying decisions based on such models. We explore the use of monotonicity constraints in classification tree algorithms. We present an application of monotonic classification trees to a problem in house pricing. In this preliminary study we found that the monotonic trees were only slightly worse in classification performance, but were much simpler than their non-monotonic counterparts.
A. J. Feelders