ادغام نظریه مجموعه راف و برنامه های کاربردی پزشکی

Medical science is not an exact science in which processes can be easily analyzed and modeled. Rough set theory has proven well suited for accommodating such inexactness of the medical profession. As rough set theory matures and its theoretical perspective is extended, the theory has been also followed by development of innovative rough sets systems as a result of this maturation. Unique concerns in medical sciences as well as the need of integrated rough sets systems are discussed. We present a short survey of ongoing research and a case study on integrating rough set theory and medical application. Issues in the current state of rough sets in advancing medical technology and some of its challenges are also highlighted.

Pawlak [1] introduced mathematical rough set theory in the early 1980’s. The theory was based on the discernibility of objects. Rough set theory provides systems designers with the ability to handle uncertainty. If a concept is ‘not definable’ in a given knowledge base, rough sets can ‘approximate’ with respect to that knowledge. From a medical point of view, the attribute-value boundaries are usually vague. In actual situations, physicians diagnose a patient and decide what is the best way to cure them. To apply rough sets to medical data and imitate this ability, many issues in rough set theory are raised [2]. For example, discretization is necessary, whether uncertainty is subjective or objective, and medical attribute values lead to difficult situations for rough set-based medical applications. These issues are also discussed by [3]. They pointed out that rough sets offer algorithms with polynomial time complexity and space complexity with respect to the number of attributes and examples. They also note that the advantages of the rough sets methodology consist of: (i) the basic tools are lower and upper approximations of the concept (which are well-defined sets) and (ii) rough sets methodology is computed directly from input data.

The other challenging open problems for rough sets are: hypothesis building, how to partition the data into appropriate distributed subsets, transformation, updating the number of examples, and attribute and system extensibility.