基于LPP-RF的基因表达数据分类  

Gene expression data classification based on LPP-RF

作　　者：杨浩宇 南晓斐[1] 柴玉梅[1] YANG Hao-yu;NAN Xiao-fei;CHAI Yu-mei(School of Information Engineering,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]郑州大学信息工程学院,河南郑州450001

出　　处：《计算机工程与设计》2018年第2期563-569,共7页Computer Engineering and Design

摘　　要：为提高随机森林算法在基因表达数据分类方面的性能,提出基于局部保持映射的随机森林算法。对每棵决策树的所有节点,通过局部保持映射,将节点数据映射到新的属性空间中,选取第一个属性为最佳分裂属性。与传统随机森林算法相比,该算法缩短了决策树的构建时间,增加了决策树之间的差异性,明显提高了分类准确率。在9个标准基因表达数据集的对比实验结果表明,该算法性能优于传统随机森林算法,对基因表达数据中类不平衡导致正类样本准确率低的问题有一定改善。To improve the performance of the random forest algorithm in the classification of gene expression data,a random fo-rest algorithm based on locality preserving projections was proposed.For all nodes of each decision tree,through the locality preserving projections,the node data were mapped to the new attribute space and the first attribute was selected as the best splitting attribute.Compared with the traditional random forest algorithm,the algorithm significantly reduces the construction time of the decision tree,increases the difference between the decision trees,and significantly improves the classification accuracy.By comparing experimental results in nine standard gene expression data sets,the proposed algorithm has better performance than the traditional random forest algorithm,and the low accuracy of positive samples caused by class imbalance of gene expression data is improved.

关 键 词：随机森林 决策树 基因表达数据 局部保持映射 分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]