Heuristic Genetic Algorithm for Discretization of Continuous Attributes in Rough Set Theory

1 Introduction Intelligent information processing is a research hotspot in information science. However, knowledge acquisition is a bottleneck in intelligent systems. Pro- posed by Pawlak in 1982[1], the rough set theory (RS theory) is based on a classification mechanism and regards knowledge as partition over data using equi- valence relationships in a given domain. The RS is a tool to deal with expressing, studying and reasoning of incomplete data and imprecise knowledge, which has been wi…

Heuristic Genetic Algorithm for Discretization of Continuous Attributes in Rough Set Theory

Discretization based on rough set theory aims to seek the possible minimum number of the cut set without weakening the indiscemibility of the original decision system. Optimization of discretization is an NP-complete problem and the genetic algorithm is an appropriate method to solve it. In order to achieve optimal discretization, first the choice of the initial set of cut set is discussed, because a good initial cut set can enhance the efficiency and quality of the follow-up algorithm. Second, an effective heuristic genetic algorithm for discretization of continuous attributes of the decision table is proposed, which takes the significance of cut dots as heuristic information and introduces a novel operator to maintain the indiscernibility of the original decision system and enhance the local research ability of the algorithm. So the algorithm converges quickly and has global optimizing ability. Finally, the effectiveness of the algorithm is validated through experiment.