基于知识粗糙熵的快速属性约简算法

针对基于正域的属性约简算法在约简过程中存在重复计算属性相对重要度从而导致算法效率低的问题，从属性度量和搜索策略的角度提出基于知识粗糙熵的快速属性约简算法。首先，在决策信息系统中通过引入知识距离提出知识粗糙熵以度量知识的粗糙程度；其次，利用知识粗糙熵作为属性显著度的评价标准来评估单个属性的重要程度；最后，利用属性重要度对所有条件属性进行排序，且通过属性依赖度删除冗余属性，从而实现快速约简。在六个公开数据集上将所提算法与其他三种算法在运行效率和分类精度上进行对比实验。结果表明，该算法的运行效率比其他三种算法分别提高了83.24%、28.77%和59.92%；在三种分类器中，分类精度分别平均提高了0.83%、0.63%和1.37%。因此，所提算法在保证分类性能的同时，能以更快的速度获得约简。

Fast attribute reduction algorithm based on knowledge rough entropy

In order to address the problem of low algorithm efficiency caused by repeated calculation of the relative importance of attributes in the attribute reduction algorithm based on positive region, this paper proposed a fast attribute reduction algorithm based on knowledge rough entropy from the perspectives of attribute measurement and search strategy. Firstly, it introduced knowledge rough entropy into decision information systems by incorporating knowledge distance to measure the degree of knowledge roughness. Next, it employed knowledge rough entropy as the criterion for evaluating the significance of attributes, assessing the importance of individual attributes. Finally, it ranked all attributes based on attribute importance, and eliminated redundant attributes through dependency, so as to achieve rapid attribute reduction. The proposed algorithm was compared with other three algorithms in terms of running efficiency and classification accuracy on six publicly available datasets. The results demonstrate that the proposed algorithm improves running efficiency by 83.24%, 28.77%, and 59.92% respectively compared to other three algorithms. Among the three classifiers, the classification accuracy increases on average by 0.83%, 0.63%, and 1.37% respectively. Therefore, the proposed algorithm is able to achieve attribute reduction more quickly while ensuring classification performance.