点得分平行坐标可视化分析方法  

A visual analysis method based on point-score parallel coordinates

作　　者：徐永红[1] 洪文学[1] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《燕山大学学报》2008年第5期440-444,470,共6页Journal of Yanshan University

基　　金：国家自然科学基金资助项目(60605006)

摘　　要：提出了一种多元数据的点得分平行坐标表示及可视化分析方法。该方法利用简单贝叶斯公式计算各属性值或属性值区间的频数和点得分,最后根据构建的点得分平行坐标即可进行数据集的可视化分析和未知样本的分类。将该方法应用到一个肝功异常数据集的结果表明,利用该图表示可以有力地揭示数据内在结构和发现知识,从而特别适合应用到疾病诊断等数据分析领域。A new multivariate data visualization method named point-score parallel coordinates （PSPC） is proposed in this paper. This method computes the frequency numbers and point scores through the naive Bayes formula, then the dataset can be analyzed and classified visually by the constructed PSPC. The result of this method applied to a liver disorders dataset indicates that it has the merits of exposing the structure of original data set and facilitating the discovery of knowledge, consequently it is very fit to be applied in many data analysis domains such as medical diagnostics.

关 键 词：多元数据 平行坐标 简单贝叶斯 点得分 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]