Knowledge Reduction based on Granular Computing

Knowledge reduction is NP-hard problem. And many approaches are proposed to get the minimal reduction, which is mainly based on the significance of the attributes. There are some disadvantages of the reduction algorithms at present. In this paper, we propose a novel heuristic function based on the distribution of granularity and treat it as important metric information of attributes. In the view of the granularity, we discussed the rationality of the heuristic function, and proposed a simple reduction algorithms based on the heuristic function. Finally, we verified the algorithm from the experiment.