We consider the problem of automatic construction of qualitative models by inductive learning from quantitative examples. We present an algorithm QUIN (QUalitative INduction) that learns qualitative trees from a set of examples described with numerical attributes. At difference with decision trees, the leaves of qualitative trees contain qualitative functional constraints as used in qualitative reasoning. A qualitative tree defines a partition of the attribute space into the areas with common qualitative behaviour of the chosen class variable. We describe a basic algorithm for induction of qualitative trees, improve it to the heuristic QUIN algorithm, and give experimental evaluation of the algorithms on a set of artificial domains. QUIN has already been used to induce qualitative control strategies in dynamic domains such as controlling a crane or riding a bicycle (described elsewhere) and can be applied to other domains as a general tool for qualitative system identification.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below