The Attribute Reduction Algorithms For Incomplete Information Systems

The rough set theory is an extension of the classical set theory, an important theory of process uncertain data, and an important method of search useful information by mining the data system. The classical rough set theory is a data process method based on specific space equivalence relation, and attribute reduction is an important research direction of it.In practice, we often need to process incomplete information systems. Especially in today's society, the information technology develop very fast, people can get a large number of relevant data of the study object easily. For various reasons, some attributes of the study object are often missing, and we could not find the original data. The information system for the raw data is an incomplete information system. The result of traditional data filled method is not ideal, so that study the attribute reduction based on incomplete information system has an important significance for apply theory to practice.This paper studies the attribute reduction based on incomplete information system in the theory of rough set and granular computing. The main works are as follows:1. Summarizes the granular computing's development process, status and basic concepts. Summarizes the main algorithm for rough set attribute reduction. Analyzes its limitations in practical application.2. Defines a new fuzzy metrics and introduces a new notion of fuzzy entropy based on it. On this basis, proposes a new algorithm named F* for incomplete information system attribute reduction. Achieves the F* algorithm by the data set from UCI DATA and ROSE data set, and compares with the IEARA algorithm. Through data analysis, F* algorithm has a better time superiority than IEARA algorithm.3. Defines a new approximate degree and introduces a new notion of fuzzy entropy based on it. On this basis, proposes a new algorithm named S*. Achieves the S* algorithm by the data set from UCI DATA and ROSE data set. The experiment results show that it has a better time superiority than the other two algorithms.