计算机科学与探索 ›› 2020, Vol. 14 ›› Issue (11): 1967-1974.DOI: 10.3778/j.issn.1673-9418.2001010

• 理论与算法 • 上一篇    下一篇

中心概念及其在规则提取中的应用

温馨,闫心怡,陈泽华   

  1. 1. 太原理工大学 大数据学院,太原 030024
    2. 太原理工大学 电气与动力工程学院,太原 030024
  • 出版日期:2020-11-01 发布日期:2020-11-09

Central Concept and Its Application in Rule Extraction

WEN Xin, YAN Xinyi, CHEN Zehua   

  1. 1. College of Big Data Science, Taiyuan University of Technology, Taiyuan 030024, China
    2. College of Electrical and Power Engineering, Taiyuan University of Technology, Taiyuan 030024, China
  • Online:2020-11-01 Published:2020-11-09

摘要:

决策信息系统的规则提取是数据分析的研究内容之一。形式概念分析是一种数据分析与信息处理的方法。从决策形式背景出发,定义综合概念以及中心概念,提出了一种在生成中心概念过程中进行规则提取的算法。在此过程中决策形式背景的决策属性参与整体概念的生成,在每一层生成的综合概念中去掉冗余综合概念之后,剩余的中心概念即为获得的决策规则。中心概念体现出信息系统中条件属性与决策属性之间的关系,给予概念更丰富的关联知识的含义,同时也实现了决策规则与形式概念之间的有效结合,所形成的概念Hasse图意义丰富而明确。该算法避免了传统方法中分别生成条件概念与决策概念,再根据它们之间的关系确定决策规则的过程复杂性。通过实例分析与基于UCI数据集上的对比实验说明了该算法的正确性与有效性。

关键词: 决策信息系统, 综合概念, 中心概念, 规则提取

Abstract:

Rule extraction of decision information system is one of the important research contents of data analysis. Formal concept analysis is efficient for data analysis and information processing. Based on formal concept, com-prehensive concepts and central concepts are defined in this paper, and then the algorithm for concise rule extraction is proposed. In this process, the decision attributes of the formal decision context take participate in the generation of the concepts. After removing redundant comprehensive concepts at each level, the rest parts are central concepts and they are the decision rules. The central concepts reflect the implication relationship between conditional attributes and decision attributes in information systems, and make formal concepts with richer knowledge. Meanwhile they also effectively combine decision rules with concepts together. The corresponding Hasse diagram is more meaningful than the traditional one. The proposed algorithm avoids the complexity of traditional methods which have to generate conditional concepts and decision concepts respectively and then to get rules by complex computation. Finally, the effectiveness and correctness of the algorithm are verified by case analysis and comparative experiments with UCI datasets.

Key words: decision information system, comprehensive concept, central concept, rule extraction