Matrix Computation for Rule Extraction in Incomplete Information Systems
不完备信息系统规则获取的矩阵算法

The incompleteness of information about objects may be the greatest obstruct to performing induction learning from example.In this paper,the concept of limited non-symmetric similarity relation is defined and then classical discernibility matrix is extended based on limited non-symmetric similarity relation.By taking the method of Boolean reasoning,rules are extracted directly from the incomplete decision systems without changing the size of original incomplete systems.The experiment shows that the algorithm provides precise and simple decision rules and does not affected by the missing values.