关键词: 属性值约简, 粗糙集, 数据挖掘, 知识发现, 规则提取算法, 归纳规则分类器

Abstract:

The existing attribute value reduction models are complex to implement， and the key information extracted by the models is often too concise， which affects the representation ability of the decision system. To resolve above problems， a heuristic attribute value reduction model based on certainty factor was proposed. Firstly， several attribute set tools with different properties were constructed， and the relevant theorems and proofs were shown； at the same time， a reduced information function was developed to assign values to the reduced attributes. Secondly， the certainty factor was taken as heuristic information and the strategy of bottom-up hierarchical search was adopted to construct a heuristic attribute value reduction model， and the layout path and operation process of the model were visually displayed in the form of the pseudo-codes of the program. Finally， the application and verification of the model were performed on simulation data from the existing research， the advantages， applicability， and scalability of the model were summarized and discussed. The results show that the new model is feasible and effective， easy to implement by programming； it has low requirements of data characteristics and is suitable for general expert systems；moreover， the value information extracted by the new model is diverse and concise with strong generalization， and does not lose the key information of the decision system.

Key words: attribute value reduction, rough set, data mining, knowledge discovery, rule extraction algorithm, inductive rule classifier