Abstract:
In many Data Mining applications the attributes are continuous (continuous input variables) and the decision values are a finite set of symbols (multi-valued output variables). In this paper we give an efficient algorithm for the minimization of Sum of Continuous-Input AND/OR implicants. Our expressions generalize the well-known Disjunctive Normal Forms to three-level structures with continuous inputs. They are compiled to rule-based programs. The network minimization uses the conditional graph coloring approach to solve the covering problem.