Abstract: In traditional attribute reduction algorithm, multiple increasing rows and the original information system are integrated into one information system, then the information system is divided into new equivalence class again. In order to reduce the complexity of the incremental data processing, the F-attribute significance in the information view is defined on the attribute set of information systems, and an incremental F-parallel attribute reduction algorithm based on information system is put forward. One or more new information systems are composed of multiple increasing rows for parallel computing. Compared with traditional attribute reduction, the incremental F-parallel attribute reduction can adapt to the newly added data well and avoid high time complexity of the incremental attribute reduction by using excellent heuristic information, the time efficiency is improved at the same time.

Key words: information system, decision subsystem, F-parallel reduction, mutual information, F-attribute significance

摘要： 在传统的属性约简算法中，将增加的对象和原来的信息系统整合为一个信息系统，并对这个信息系统重新划分新的等价类，为了降低处理增量式数据的复杂度，在信息系统的属性集上定义了信息论意义下的F-属性重要度，给出了一种增量式F-并行属性约简算法。该算法将增加的多行记录组成一个或若干个新的信息系统进行并行计算。与传统的方式相比，该算法能够很好地适应新加入的数据，同时通过利用优秀的启发式信息避免了增量式属性约简时间复杂度过高的问题，时间效率提高。

关键词: 信息系统, 决策子系统, F-并行约简, 互信息, F-属性重要度