基于测试代价的三支邻域属性约简算法

针对粗糙集属性约简时很少考虑属性自身的测试代价等问题，提出了一种基于测试代价的三支邻域属性约简算法。算法根据各属性在邻域分辨矩阵中出现的频次和比例来计算属性重要性，并结合属性自身的测试代价来构造性价比指标，以此指导属性的甄选。三支决策方法被用于划分属性集，为属性的约简处理提供数据支撑。在7个UCI公共数据集上进行对比实验，结果表明，该算法可得到比对比算法更小的属性约简集合，在分类精度不降低的情况下，该算法具有更少的运行时间和更小的测试代价。基于财政收入的预测应用实例进一步证明了所提算法的有效性和实用性。

Three-way neighborhood attribute reduction algorithm based on test cost

In order to address the issue of test cost being rarely considered in rough set attribute reduction, this paper proposed a three-way neighborhood attribute reduction algorithm based on test cost. The proposed algorithm calculates the attri-bute importance according to the frequency and proportion of each attribute in the neighborhood resolution matrix, and combined the test cost of the attributes to construct the the cost performance index to guide the selection of attributes. Three-way decision-making method was employed to partition attribute sets, which provides data support for the attribute reduction process. Comparative experiments were conducted on seven UCI public datasets, which demonstrate that the proposed algorithm yields a smaller attribute reduction set compared to the comparison algorithm. Moreover, the proposed algorithm exhibitd a shorter running time and lower test cost without compromising classification accuracy. Furthermore, it provided an application example based on fiscal revenue prediction to further validate the effectiveness and practicality of the proposed algorithm.