高密度遗传图谱单标记选择的随机森林法  

Single-mark selection method for high-density genetic map based on random forest variable importance score

作　　者：郑珂晖[1] 王健[2] ZHENG Kehui;WANG Jian(College of Computer and Information, Fujian Agriculture and forestry University, Fuzhou, China, 350002;College of Geography, Fujian Normal University, Fuzhou, China, 350007)

机构地区：[1]福建农林大学计算机与信息学院,福州350002 [2]福建师范大学地理科学学院,福州350007

出　　处：《福建电脑》2019年第5期21-24,共4页Journal of Fujian Computer

基　　金：福建省教育厅中青年科研项目(No.JA15160)资助

摘　　要：高密度图谱上的分子标记数量众多,属于典型的高维数据,用传统的回归分析方法难以筛选。随机森林是一种基于决策树的算法,通过对决策树进行汇总,提高了模型的预测精度,可以用来解决回归问题与分类问题。本研究采用随机森林中的变量重要性评分的方法来对高密度遗传图谱上的与性状相关的单个标记进行选择,对不同遗传率、不同群体大小的情况进行了模拟研究,每个参数组合模拟100次,计算选出标记位置的均值与标准差,统计选择正确的次数,模拟结果表明该方法是一种行之有效的方法。The number of molecular markers on the high-density map is numerous and belongs to typical high-dimensional data, which is difficult to screen by traditional regression analysis methods. Random forest is a decision tree-based algorithm. By summarizing the decision tree, the prediction accuracy of the model is improved, which can be used to solve the regression problem and classification problem. In this study, the random forest algorithm is applied to the marker selection of high-density maps. Through simulation studies, it is found that the random forest algorithm has a better effect on single genetic marker selection on high-density genetic maps.

关 键 词：随机森林 遗传标记 高密图遗传图谱 标记选择 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]