Abstract: Medical data mining can generate effective knowledge rules from medical record database.It is found that the precision and speed of medical diagnosis are unsatisfied due to large-scale repeating data and redundant attributes in medical data during practical applications.To solve the problem，this paper establishes relation of rough set theory based on information view and SQL language，and conditional information entropy attributes reduction algorithm based on SQL language is put forward.Experiments show that the algorithm can be easily realized taking advantage of database query language and is more efficient.It establishes a method by which rough set theory is widely used in material medical data mining.