基于随机森林算法的特征选择及在fMRI数据中的应用  

Feature Selection Based on Random Forest Algorithm and the Applications in fMRI Data

作　　者：袁晓龙[1] 梅雪[1] 黄嘉爽 杨骥[1] 

机构地区：[1]南京工业大学自动化与电气工程学院,江苏南京211816

出　　处：《微电子学与计算机》2014年第8期132-135,共4页Microelectronics & Computer

基　　金：国家自然科学基金(51205185);2012年度江苏省"青蓝工程"中青年学术带头人项目

摘　　要：fMRI数据是典型的高维小样本数据,如何从高维数据中提取和选择重要的特征是正确分类识别的关键.结合fMRI数据自身特点,提出了一种应用随机森林算法进行特征选择的方法,以随机森林分类精度为准则函数对特征进行重要性度量的方式实现特征选择.将本方法应用于健康者和精神分裂症患者的识别中,通过计算每个特征对分类的贡献度,优选出贡献度大的特征用于分类识别,同时根据重要特征的序号定位到相应脑区,给临床诊断提供客观参照.实验结果表明,该方法具有较好的效果.fMRI data is typical small sample data with high-dimension ,how to extract and select important features from high-dimensional data is the key for classification and recognition .Combined with the characteristics of fMRI data ,a method of feature selection using Random Forest algorithm is proposed in this paper .To achieve the feature selection ,Random Forest classification accuracy as a criterion to estimate the importance of features .The method was applied to healthy controls and patients with schizophrenia recognition ,by calculating the contribution of each feature for classification to choose the features with greater contribution for classification and recognition ,and locate to the corresponding brain regions according the order numbers of important features ,providing an objective reference for clinical diagnosis .Experimental results show this method has a better performance .

关 键 词：分类识别 特征提取与选择 随机森林 fMRI数据 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]