An incremental attribute reduction approach based on knowledge granularity with a multi-granulation view.

Dynamic updating of attribute reduction is a key factor for the success of rough set theory since many real data vary dynamically with time. Though many incremental methods for updating reduct have been proposed to deal with a dynamically-varying data set and has attracted much attention. However, it is hard to update reduct when the large-scale data vary dynamically. To overcome this deficiency, in this paper, we develop an attribute reduction algorithm with a multi-granulation view to discover reduct of large-scale data sets. Then, incremental mechanisms for knowledge granularity are introduced and two corresponding incremental approaches for updating reduct are developed when many objects are varied in a large-scale decision table with a multi-granulation view. Finally, experiments have been run on six data sets from UCI and the experimental results show that the proposed incremental algorithm with a multi-granulation view can achieve better performance for large-scale data sets.