摘要
结合信息可视化与机器学习技术,提出一种基于多元数据平行坐标图表示的贝叶斯可视化分类方法。该方法基于类条件概率密度估计对平行坐标图表示进行优化,最后对变换后的各变量值加权求和,用贝叶斯法则分类。这种方法通过平行坐标来使不可见的数据和算法变得可见,从而易于利用专家领域知识,分类结果容易理解,特别适合应用到疾病诊断等医学领域的模式识别问题。
This article combines the techniques of information visualization and machine learning to bring forward a Bayesian visual classification method based on parallel coordinates graphical representation of multivariate data.This method optimizes the parallel coordinates representation by class-conditional probability density estimations, then the transformed point score values are weighted and summed up to apply the Bayesian classification rules. This method makes the invisible data and algorithms visible by using parallel coordinates, consequently domain experts＇ knowledge is easier to be utilized,classification results are more interpretable so as to qualify it as an effective tool for some pattern recognition tasks such as medical diagnostics.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第25期166-169,188,共5页
Computer Engineering and Applications
基金
国家自然科学基金No.60605006
No.60671025~~
关键词
模式识别
信息可视化
多元数据
平行坐标
简单贝叶斯
pattern recognition
information visualization
multivariate data
parallel coordinates
naive-hayes
