序决策系统下基于图顶点最小覆盖的属性约简
Attribute Reduction Based on Minimum Cover of Graph Vertices in Ordered Decision System

现如今的互联网时代，数据维度灾难性增长，如何从高维数据中提取有用信息成为一大难题。属性约简是数据预处理的重要步骤之一，能够减少属性维度和计算复杂度，提高分类性能和可解释性。传统的属性约简方法主要基于信息论、统计学或启发式算法，存在不足之处。本文提出了一种基于图顶点最小覆盖的序决策系统属性约简方法，利用图来建模属性之间的依赖关系，使属性约简算法和图论知识相结合。实验结果表明，本文方法在多个数据集上具有较好的约简效果和分类性能，具有良好的可解释性和可视化效果。
In today’s Internet era, the data dimension has grown more dramatically, and how to extract useful information from high-dimensional data has become a big problem. Attribute reduction is one of the important steps of data preprocessing, which can reduce the attribute dimension and computa-tional complexity, and improve the classification performance and interpretability. Traditional at-tribute reduction methods are mainly based on information theory, statistics or heuristic algorithms, which are shortcomings. In this paper, we propose a method based on the minimum cover-age of graph vertices to model the dependencies between properties and combining the attribute reduction algorithm and graph theory knowledge. Experimental results show that the present method has good reduction and classification performance on multiple datasets, with good inter-pretability and visualization.