rpartScore-package(rpartScore)
rpartScore-package()所属R语言包：rpartScore
Classification trees for ordinal responses
分类树为序反应
译者：生物统计家园网 机器人LoveR
描述----------Description----------
This package contains functions that allow the user to build classification trees for ordinal responses within the CART framework.<br> The trees are grown using the Generalized Gini impurity function, where the misclassification costs are given by the absolute or squared differences in scores assigned to the categories of the response.<br> Pruning is based on the total misclassification rate or on the total misclassification cost.
此套件包含的功能，允许用户建立分类序反应的CART框架内的树木。<BR>的树木生长使用广义的的基尼杂质功能，误判成本的绝对或平方差的分数的响应的类别是根据<br>文章修剪的总错误率或总误判成本。
Details
详细信息----------Details----------
This package contains functions that allow the user to build classification trees for ordinal responses within the CART framework.<br> It is assumed that a set of numerical scores has been assigned to the ordered categories of the response. <br> Two splitting functions are implemented, both based on the generalized Gini impurity function. They use the absolute and the squared differences in scores, respectively, as misclassification costs. <br> In order to select the optimal tree size, pruning can be performed, using two different measures of prediction performance: the total misclassification rate or the total misclassification cost.<br> This package requires the rpart package. The main function in this package is rpartScore. The use of this function is almost the same as the rpart function. The main difference is the presence of two arguments (split and prune) instead of the method argument.<br> The argument split controls the splitting function used to grow the classification tree, by setting the misclassification costs equal to the absolute ("abs" - default option) or to the squared ("quad") differences in scores. <br> The argument prune allows the user to select the prediction performance measure used to prune the classification tree, and can take two values: "mr" (total misclassification rate) or "mc" (total misclassification cost - default option).
此包中包含的功能，允许用户建立分类树的CART框架内的序反应。<br>这假定有序类别的响应已被分配到的一组数值的分数。参考两个分裂功能的实现，无论是基于广义的基尼杂质功能。他们使用绝对的平方，成绩的差异，分别作为误分类成本。 <br>为了选择最优树的大小，可以进行修剪，使用两种不同的预测性能的措施：总的错误率或误判成本。<br>此包需要rpart包。在此套件的主要功能是rpartScore。此函数的使用作为rpart功能几乎是相同的。主要的区别是存在的两个参数（split和prune），而不是method的说法。<BR>参数split控制的分离功能增长分类树，通过设置错误分类成本等于绝对（"abs" - 默认选项）的平方（"quad"），成绩的差异。参考参数prune允许用户选择的预测性能指标用于修剪的分类树，并可以取两个值："mr"（总错误率）或"mc"（总误判成本 - 默认选项）。
（作者）----------Author(s)----------
Giuliano Galimberti, Gabriele Soffritti, Matteo Di Maso
Maintainer: Giuliano Galimberti <giuliano.galimberti@unibo.it>
参考文献----------References----------
Journal of Statistical Software, 47(10), 1-25. URL http://www.jstatsoft.org/v47/i10/.
Technical Report 61, Section of Biostatistics, Mayo Clinic, Rochester. URL http://www.mayo.edu/hsr/techrpt/61.pdf.
参见----------See Also----------
rpart
rpart
实例----------Examples----------
require("rpartOrdinal")
data("lowbwt")
lowbwt$Category.s <- ifelse(lowbwt$bwt <= 2500, 3,
ifelse(lowbwt$bwt <= 3000, 2,
ifelse(lowbwt$bwt <= 3500, 1, 0)))
T.abs.mc <- rpartScore(Category.s ~ age + lwt + race + smoke +
ptl + ht + ui + ftv, data = lowbwt)
plotcp(T.abs.mc)
T.abs.mr <- rpartScore(Category.s ~ age + lwt + race + smoke +
ptl + ht + ui + ftv, data = lowbwt, prune = "mr")
T.quad.mc <- rpartScore(Category.s ~ age + lwt + race + smoke +
ptl + ht + ui + ftv, split = "quad", data = lowbwt)
T.quad.mr <- rpartScore(Category.s ~ age + lwt + race + smoke + ptl + ht +
ui + ftv, split = "quad", prune = "mr", data = lowbwt)
转载请注明:出自 生物统计家园网(http://www.biostatistic.net)。
注：
注1：为了方便大家学习，本文档为生物统计家园网机器人LoveR翻译而成，仅供个人R语言学习参考使用，生物统计家园保留版权。
注2：由于是机器人自动翻译，难免有不准确之处，使用时仔细对照中、英文内容进行反复理解，可以帮助R语言的学习。
注3：如遇到不准确之处，请在本贴的后面进行回帖，我们会逐渐进行修订。
|
发表于 2012-9-27 23:47:42
