基于小波去噪和改进的FCM算法的基因表达数据分析  

Analysis of gene expression data based on wavelet denoising and improved fuzzy c-means algorithm

作　　者：陈刚[1] 陆媛[1] 杨慧中[1,2] 

机构地区：[1]江南大学教育部轻工过程先进控制重点实验室,江苏无锡214122 [2]上海市电站自动化技术重点实验室,上海200072

出　　处：《计算机与应用化学》2011年第7期951-953,共3页Computers and Applied Chemistry

基　　金：国家自然科学基金资助项目(60674029);江苏省高技术研究项目(BG2006010);上海市科学技术委员会的资助(09DZ2273400)

摘　　要：模糊c均值算法是一种局部搜索迭代法,易陷入局部最小解,而且算法未考虑样本对聚类的贡献程度。针对传统的模糊c均值(FCM)算法的不足和基因表达数据高噪声的特点,提出了一种基于小波变换和改进的FCM聚类模型,最后将该模型应用于白血病基因数据分析。根据Xie-Beni指数,在没有先验知识的条件下,确定了最佳聚类个数。为了体现文中提到的算法对样本聚类的准确性,本文分别采用传统的FCM聚类算法和分层聚类的方法在同样的试验条件下进行试验。样本聚类的结果表明:该方法能得到高准确度的样本分型结果。Fuzzy c means algorithm is an iteretive local search method, easily falling into local minimum solution, and the clustering algorithm does not consider the contribution of th,： sample. Aiming at the deficiencies of traditional FCM algorithms and the feature of much noise in gene expression data, this paper proposed a new clustering model based on multi-resolution decomposition and weighted possibilistic fuzzy cluster analysis. The model is applied to th,.~ gene expression data of leukaemia. According to Xie-Beni index, in the absence of prior knowledge of the conditions, the optimum number of clusters is made. To reflect the sample clustering algorithm accuracy mentioned in the text, this paper uses both the traditional FCM clustering algorithm and the hierarchical clustering method in the same experimental conditions to take a test. The results of clustering show that the precise clustering of disease types can be achieved with method.

关 键 词：基因表达数据 小波变换 改进的FCM聚类 样本聚类 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]