A New Rule Extraction Method and Algorithm by Rough Sets Based on a Rule Space

Rough sets are often used for extracting rules from categorical data sets with condition and decision attributes. However, the conventional method to extract these rules has difficulty in penetrating the extracting processes and in examining the validity of the results. Then, we have reviewed and arranged the conventional method in a rule space which consists of atom rules and gives an intelligible interpretation of the conventional method. In this paper, we propose a new rule extraction method which is located in the middle of the lower and upper approximation method of the conventional method. We also provide a new algorithm for extracting rules in the rule space, implement that algorithm in computer software, and examine the efficiency and the merits and demerits between the software developed for the conventional method and our improved method.