摘要案例特征属性权重的选取直接影响到案例检索的精度,因此本文针对案例检索中权重确定方法进行了比较研究。首先研究了案例检索中各种权重确定的方法的基本理论和概念,并在此基础之上总结出各种权重确定方法针对的难题、存在的问题以及计算权值的步骤。接下来着重分析了相似粗糙集权值算法、基于知识熵的案例特征属性权重算法和特征属性自动学习算法这三种权重确定方法,采用一个工程项目风险分类系统的案例,分别运用上述三种权重确定方法给此案例的特征属性赋予权重,最后运用相似度测量来检验三种权值的正确性,为了避免一种相似度测量方法对结果造成的偶然性,本文选择了K-NN和TF-KNN这两种相似度算法来分别对得到的权重进行案例检索的检验,案例检索的有效性和准确性是检验的标准,所以本文从查全率和查准率两个方面来验证综合权值计算过程等进行分析,得出三种权重确定方法中最有效的方法。65185
关键词 案例检索 权重 相似度 案例推理
毕业论文 外 文 摘 要
Title Study on comparing of the weight determination methods in case retrieval
Abstract
The selection of the characteristic attribute weights directly affects the accuracy of case retrieval, so this paper is to study on comparing of the weight determination methods in case retrieval. Firstly, it studied the basic concept and theories of various weight determination methods in case retrieval, and summarizes the advantages of these weight determination methods, existing problems and the steps of calculating the weight values. Then emphatically analyzed three methods of weight determination, including the SRS algorithm, methods based on knowledge entropy and the automatic learning characteristic attributes algorithm. Thirdly, this paper used a case of engineering project risk classification system, then give weight values to he characteristic attributes of this case using the above three methods respectively. Finally, using similarity algorithms to test the correctness of the weights. In order to avoid the contingency caused by one similarity measure, this paper chooses two similarity algorithms K-NN and TF-KNN to test the weights we got respectively. The validity and accuracy is the test standard of the case retrieval, so this article from the recall and precision rate two aspects to validate comprehensive weight calculation processes for analysis, the three kinds of weight determination methods in the most effective way.
Keywords case retrieval weights similarity case-based reasoning
目次
1 引言 1
1.1 研究背景 1
1.2 研究目的及意义 1
1.3 案例推理 2
1.3.1 案例推理的定义 2
1.3.2 案例推理的流程 3
1.4 案例检索 4
1.4.1 案例检索的定义 4
1.4.2 案例检索的流程 4
1.4.3 案例检索的方法 5
2 权重确定的方法 6
2.1 传统的定权方法 6
2.2 优化后的定权方法 6
2.3 权重确定方法存在的问题 9
3 权重确定的算法内容 10