基于决策支持向量机的产品设计知识文档分类研究  

Product design knowledge document categorization based on decision tree & SVM

作　　者：车君华[1] 冯毅雄[1] 谭建荣[1] 王云[1] 

机构地区：[1]浙江大学CAD&CG国家重点实验室

出　　处：《计算机集成制造系统》2007年第5期891-897,共7页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(60573175;50475072;50505044);国家973计划基金资助项目(2004CB719400)~~

摘　　要：为准确地定位大量旧的产品设计知识文档信息,构建产品设计知识库,提出了基于决策树和支持向量机的产品设计知识文档分类方法。通过对产品设计知识文档的分词处理、特征抽取、特征合并和向量表达,为决策支持向量机方法提供了样本数据预处理。在分类算法中,为了改善样本的处理效率,通过二叉决策树依次构建多个支持向量机对文档样本进行分类。同时为了保证样本分类的准确性,在最后一个分类添加一个分类器,再次对样本进行判断,避免出现因前期分类器训练不足而产生的误判。最后,通过分类性能评价理论的验证和企业信息化项目的应用,证明了该方法的有效性和实用性。To precisely allocate the massive old Product Design Knowledge Document （PDKD） and construct the knowledge base of product design, the PDKD categorization method based on decision tree and Support Vector Machine （SVM） was put forward. For the application of the PDKD categorization method, the preprocessing of sample data was obtained by word segmentation, characteristics extraction, characteristics combination and vector expression of the PDKD. In the categorization algorithm, multiple SVMs were constructed in turn with binary decision tree to categorize the sample data to improve the categorization processing efficiency. Then the last category was judged to avoid the former categorizer＇s misjudgment due to lack of trainings, thus the categorization processing accuracy was ensured. Finally, the validity of the categorization method was proved by the verification of the categorization evaluation theory. The feasibility and practicality of the above-mentioned method were approved by an application case of the enterprise information project.

关 键 词：设计知识 支持向量机 决策树 分类 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]