应用有效信息比率的离散化算法  

Discretization Algorithm Based on Effective Information Ratio

作　　者：诸文智[1] 王靖程[1] 张彦斌[1] 贾立新[1] 

机构地区：[1]西安交通大学电气工程学院,西安710049

出　　处：《西安交通大学学报》2011年第4期12-17,共6页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(61005058);教育部高等学校博士学科点专项科研基金资助项目(20090201110019;20070698059)

摘　　要：针对目前离散化信息量度无法准确表征数据离散后有效分类信息量的问题,提出了一种基于有效信息比率的离散化算法.在构建离散化方案相依表的基础上,分析了离散区间内类属性分布与分类信息蕴含量间的关系,并根据类属性分布信息引入有效信息比率,用于表征各离散区间内有效分类信息量.然后,依据离散化方案的离散区间数及其有效信息比率,设计出表征离散化方案划分质量的离散化评价指标,从而提高了数据的离散化效果.仿真实验和实际应用的结果表明,该算法离散化后在有效分类信息量和分类预测精度上高于主流基于信息论的离散化算法.Since the current information measures of discretization can not accurately reflect the degree of the effective class information in the discretized dataset, a discretization algorithm based on effective information ratio is presented, and a contingency table of corresponding discretization scheme is constructed. According to the analyses of the relationship between the class distribution and the remaining class information, the effective information ratio based on the class distribution is analyzed to indicate the degree of effective information in each discretized interval. An improved discretization criterion is generated to evaluate the quality of the discretization scheme following the number of discrete internal and the effective information ratio. The simulation and applications illustrate the more effective class information and the higher classification accuracy than the other information-based solutions.

关 键 词：离散化 信息 有效信息比率 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]