中医察目望神规则挖掘的关键技术  

Rules Mining Research on the Eye Features Computation in the Spirit Diagnosing of TCM

作　　者：郭锋[1,2] 李绍滋[1,2] 戴莹 周昌乐[1,2] 林颖[1,2] 

机构地区：[1]厦门大学信息科学与技术学院,福建厦门361005 [2]福建省仿脑智能系统重点实验室(厦门大学),福建厦门361005 [3]岩手县大学软件与信息科学部

出　　处：《厦门大学学报（自然科学版）》2011年第3期525-530,共6页Journal of Xiamen University：Natural Science

基　　金：国家自然科学基金项目(60672018);深圳市科技计划基础研究项目(JC200903180630A)

摘　　要：中医客观化中有一个很重要的问题是根据量化的人体特征推导出人体的状态.提出了一种用于中医察目望神客观化中从眼部特征推导出人体"神"的状态的规则挖掘方法.首先给出了视频采集方法,接着使用类-属性依赖最大化方法(CAIM)对眼部特征数据进行离散化形成规则挖掘中的属性区间,然后使用云模型进行关联规则挖掘得到大量候选规则,再对3种互补的规则裁剪方法进行修改和组合用于候选规则的整理,并形成最终的规则集合.利用交叉验证法检验规则挖掘的效果,得到了93%的平均精确度,达到了很好的效果.Quantitative features reflecting the human body′s appearance are very helpful in diagnosing human′s health state in Traditional Chinese Medicine（TCM）.This paper presents a novel application of eye features computation in the Spirit diagnosing,for which rules describing ＂the Spirit＂ state are mined by the quantitative features regarding the human eyes.With videos capturing the eye condition during a short time,a set of eye features are extracted.On the basis of it,attribute intervals of the eye feature space are generated by the CAIM.Then,a various of candidate rules are mined by the association rule mining based on the Cloud model.Finally,three complementary rule-pruning methods are modified and combined to trim the boring candidate rules.The cross validation test for mined rules reaches the average accuracy of 93%,which shows the good performance of the proposed method.

关 键 词：察目望神客观化 云模型 类-属性依赖最大化方法 规则裁剪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]