基于支持向量机的拟南芥基因表达数据分析

摘要

For the problem of Arabidopsis root gene expression analysis, this paper presentsa new algorithm of multi-class Support Vector Machines (SVMs) , which is based onlearned distance measure. Because of speciality of this problem, a distance measureis learned by minimizing Leave-one-out (LOO) error of 4-class SVMs, and some genesbelong to other classes are determined, then 5-class SVMs is constructed to classify thetotal genes. Experiments prove the effective of our method compared with traditionalclustering methods.%针对拟南芥根部基因表达数据分析的问题,本文提出了一种新的基于距离度量学习的支持向机多分类算法.鉴于此问题的特殊性,本文通过最小化4分类机的LOO误差来求得一个恰当的距离度量.并在此度量下找到若干个属于第5类(其它类)的训练点,从而构造出一个5分类机用来对所有基因分类.实验验证了此算法的可行性,并且比基因表达分析中传统使用的聚类方法更有效.