一种发现函数依赖集的方法及应用

A Method for Discovering Functional Dependencies and Its Application

介绍了一种发现最小函数依赖集的方法.这种方法基于一致集的概念,根据一致集导出最大集及其补集,然后生成最小非平凡函数依赖集.通过使用带状划分数据库减少求一致集的运算次数,使用逐层求精的算法来计算最小非平凡函数依赖集的左部.其结果可用于数据库的重新组织和设计、属性约简、聚类、关联规则提取等知识发现工作中.

An efficient method is introduced for discovering minimal functional dependencies from large database. It is based on the concept of agree sets. From agree sets, maximal sets and its complements are derived, and all minimal non-trivial functional dependencies can be generated. The computation of agree sets can be decreased by using stripped partition database. A levelwise algorithm is used for computing the left hand sides of minimal non-trivial functional dependencies. This method can be used to attribute reduction, clustering and mining associate rules, etc. in knowledge discovery as well as reorganization and design of databases.