关于粗糙集的综合评价方法研究与应用

【作者】 张帆；

【导师】 李其深；

【作者基本信息】 西南石油大学 ， 应用数学， 2015， 硕士

【摘要】 粗糙集理论是一种基于数据进行的客观、定量的数据挖掘方法,具有知识分类、描述不确定性、量化属性依赖性和重要性、属性约简等功能特点。基于这些特点能使其与综合评价方法进行适当结合来得到更合理真实的评价结果。本文在系统地将粗糙集理论引入并应用到综合评价的过程中,主要进行了如下工作：基于粗糙集理论从三个方面研究了综合评价过程中的属性权重计算方法。在确定属性客观权重方面,通过基于信息量的属性重要性、基于知识粒度的属性重要性、基于分辨矩阵以及基于属性依赖性这四种方法对客观权重赋值；在确定属性主客观组合权重方面,利用专家经验知识得出主观权重,引入经验因子合成得到属性的主客观组合权重；在确定多种方法的组合权重方面,利用粗糙集和信息熵首先计算各方法的权重,进而与各方法确定的属性权重进行合成得到属性的方法组合权重。建立粗糙集与其它评价方法结合的综合评价模型。基于粗糙集的综合评价中,对于一般问题可以总结一个程序化的评价流程。在评价的过程中可以巧妙运用分辨矩阵对评价方法进行筛选。在与其它传统评价方法融合的过程中,粗糙集利用自身特点可以使得结果更加客观合理。在粗糙集与模糊集理论结合的过程中,将模糊集的聚类能力发挥到粗糙集对于分类的理解上；在粗糙集与群决策结合的过程中,将粗糙集的客观性与群决策的相互影响结合得到更真实的结果；在粗糙集与层次分析法结合的过程中,将粗糙集的定量分析与层次分析法的定性分析结合来还原每一层次的真实排序。更多还原

【Abstract】 Rough set theory is a kind of data mining methods. In the paper we combined the rough set theory with the comprehensive evaluation methods. What the main works we have done are as follows:We have studied the methods which can get the property rights in the process of comprehensive evaluation based on the rough set theory from three aspects. In the aspect of objective weights. There are four methods which are the importance of information,the importance of knowledge granularity,the discernibility matrix and the attribute’s dependency to get the weights. In the aspect of comprehensive weights. Comprehensive weights were obtained from the combination which by experience factor of subjective weights by experts’ experience and objective weights;In the aspect of method weights. It used rough set and information entropy first to determine the weight of each method,and then admix it with the weight of each property by each method to get the the final combined weights.We also have Combined rough set with other evaluation methods to established the comprehensive evaluation model. We have summarized a programmatic evaluation process based on rough sets. In the evaluation process,the discernibility matrix can be use to remove unnecessary methods. Rough set make the results more objective and reasonable in the process of integration with other traditional evaluation methods. Fuzzy clustering can be use to classification;The combination of rough set and group decision will make more realistic results;AHP based on rough set will make the order of every level more reasonable.更多还原