高维数据中有限混合模型特征选择的应用研究

The Application of Finite Mixture Model Feature Selection in High-dimensional Data

在计算速度飞速发展的今天,统计学家们已经开始研究超高维和复杂的数据,尤其在生物信息学方面更显突出.相应的统计问题表现在观测数据中有很多的特征来自不同的子总体.在每个子总体中选择数量恰当的特征(变量)是得到好的模型预测的关键.本文基于稀疏正态线性模型(FMSL)理论,对基因转录组数据进行了分析,筛选出重要的特征变量.

In today's computing speed rapid development,scientists have begun to study the ultra high dimensional and complex data,especially more prominent in bioinformatics. The corresponding statistical problems in attention to observation data in the overall. There are a lot of features from different sub-population. Choosing appropriate features in each sub-population(variable) is the key to get good models predict. This article is based on finite mixture sparse normal linear model(FMSL) theory,analyses the gene transcription of data,choose the important characteristics of a variable.