基于随机森林的基因表达数据分类强度研究-模式识别与智能系统专业论文.docxVIP

摘要摘要 摘要 摘要 利用DNA微阵列技术产生的基因表达谱数据，进行疾病诊断、治疗、药物 研制和药物筛选是当前的一个研究热点。由于疾病诊断和药物研制直接关系到人 类的健康，因此研究基因表达数据的分类结果的可信度有着特别重要的意义。 针对基因表达谱数据的分类可信度问题，本文利用随机森林的分类强度对其 研究，其主要工作有： 1．详细介绍了基因表达谱数据分类问题的研究现状和随机森林技术的算法 原理。 2．针对随机森林现有工具包的缺点，开发并实现了C语言版的随机森林工 具包，并给出了MATLAB的接口，使其亦能在MATLAB平台运行。该工具包 简便实用，有利于随机森林的应用推广。 3．针对基因表达谱数据的分类可信度问题，提出了基于随机森林的票数差 和Proximity矩阵接近度之差的两种可信度度量。四个基因表达谱数据集的实验 显示，两种可信度量都能准确表征了基因表达谱数据分类的可信度，可信度大的 样本很少出现误判现象，而可信度小的样本常会出现误判现象，表明了这两种可 信度度量的有效性。 关键词：基因表达谱数据；可信度；随机森林；分类强度 AbstractAbstract Abstract Abstract Using DNA microarray technology produces gene expression data to diagnosis， treatment,drug development and drug screening is currently a research focus．As the disease diagnosis and drug development is directly related to human health，the study has classified the results ofthe credibility ofparticular importance． Credibility for the classification of gene expression data,the paper strength of the classification using Random Forests，the main work： I．Details the gene expression data classification problems Research and Random Forests algorithm technology works． 2．Random Forests package for the shortcomings of existing tools，develop and implement a C language version of the Random Forests tools package，and gives the MATLAB interface，it can also rim in the MATLAB platform．The tools package is simple and practical，will help to promote the application of Random Forestss． 3．Classification for gene expression data credibility problem，proposed a number of votes based on random forest proximity poor and Proximity matrix credibility of the difference between the two measures．A set of four experiments showed that gene expression data，both the amount of credibility can accurately characterize the gene expression data classification credibility,the credibility of a large sample of rare false positive phenomenon．and the credibility of the small sample of the phenomenon of false positives often OCCUr,indicating the effectiveness of these two credibility Keyword