基于阿尔茨海默病的基因表达数据改进的聚类方法
发布时间：2018-07-18 10:16
【摘要】：阿尔茨海默病以其高发病率和无法治愈的特点成为老年人的第四大“健康杀手”.迄今为止,尚不明确阿尔茨海默病的发病机理.随着基因芯片技术的发展,基因表达数据的聚类分析方法应用到阿尔茨海默疾病的研究中.为了挖掘阿尔茨海默病的生物信息,本文提出了一种改进的聚类算法.通过实验分析,得到了阿尔茨海默病基因表达数据的特征.当数据呈现出线性相关的特征时,对高维数据进行降维处理并判断聚类趋势,得到数据具有明显的聚类趋势,然后选择用聚类的方法挖掘基因表达数据的潜在信息.但是传统的聚类方法需要事先确定出分类个数,而主观的参数选取难以将大量的数据准确地进行聚类,而且这个参数的选取直接影响了实验数据的分类结果,使得聚类效果缺乏客观性.因此,本文提出了将曲率最大处作为分类判据的无监督聚类方法,并且给出了分类判据δ.在基于拟合方法的基础上,找出四种不同患病程度下聚类的阈值和聚类结果.最后得出本文的聚类效果优于其他聚类方法的效果.同时实验结果表明,这种改进的聚类方法较优,并且简捷、具有可行性.
[Abstract]:Alzheimer's disease, with its high incidence and incurable characteristics, has become the fourth largest health killer in the elderly. So far, the pathogenesis of Alzheimer's disease remains unclear. With the development of gene chip technology, cluster analysis of gene expression data has been applied to the study of Alzheimer's disease. In order to mine the biological information of Alzheimer's disease, an improved clustering algorithm is proposed in this paper. The characteristics of Alzheimer's disease gene expression data were obtained by experimental analysis. When the data show linear correlation characteristics, the high dimensional data is reduced and the clustering trend is judged, and the clustering trend is obtained, and then the potential information of gene expression data is mined by clustering method. But the traditional clustering method needs to determine the number of classification in advance, but the subjective parameter selection is difficult to cluster a large number of data accurately, and the selection of this parameter has a direct impact on the classification results of experimental data. The clustering effect lacks objectivity. Therefore, an unsupervised clustering method using the maximum curvature as the classification criterion is proposed, and the classification criterion 未 is given. On the basis of fitting method, the threshold and clustering results of four kinds of diseases were found out. Finally, it is concluded that the clustering effect of this paper is better than that of other clustering methods. The experimental results show that the improved clustering method is simple and feasible.
【学位授予单位】：四川师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.13
本文编号：2131592
[Abstract]:Alzheimer's disease, with its high incidence and incurable characteristics, has become the fourth largest health killer in the elderly. So far, the pathogenesis of Alzheimer's disease remains unclear. With the development of gene chip technology, cluster analysis of gene expression data has been applied to the study of Alzheimer's disease. In order to mine the biological information of Alzheimer's disease, an improved clustering algorithm is proposed in this paper. The characteristics of Alzheimer's disease gene expression data were obtained by experimental analysis. When the data show linear correlation characteristics, the high dimensional data is reduced and the clustering trend is judged, and the clustering trend is obtained, and then the potential information of gene expression data is mined by clustering method. But the traditional clustering method needs to determine the number of classification in advance, but the subjective parameter selection is difficult to cluster a large number of data accurately, and the selection of this parameter has a direct impact on the classification results of experimental data. The clustering effect lacks objectivity. Therefore, an unsupervised clustering method using the maximum curvature as the classification criterion is proposed, and the classification criterion 未 is given. On the basis of fitting method, the threshold and clustering results of four kinds of diseases were found out. Finally, it is concluded that the clustering effect of this paper is better than that of other clustering methods. The experimental results show that the improved clustering method is simple and feasible.
【学位授予单位】：四川师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.13
【参考文献】
相关期刊论文 前6条
1 付如意;黄静;胡本琼;庞朝阳;;应用阿兹海默症基因表达数据对比2种层次聚类方法[J];四川师范大学学报(自然科学版);2015年06期
2 宋丽红;;K-均值聚类的Matlab仿真设计[J];实验技术与管理;2010年10期
3 孟文超;王纯莹;;老年痴呆研究进展[J];军医进修学院学报;2008年05期
4 岳峰;孙亮;王宽全;王永吉;左旺孟;;基因表达数据的聚类分析研究进展[J];自动化学报;2008年02期
5 张建萍;刘希玉;;基于聚类分析的K-means算法研究及应用[J];计算机应用研究;2007年05期
6 胡本琼,张先迪,庞朝阳;利用图论设计图像压缩中的向量量化聚类算法[J];四川师范大学学报(自然科学版);2005年03期
，本文编号：2131592
本文链接：https://www.wllwen.com/kejilunwen/jiyingongcheng/2131592.html
最近更新
教材专著
