基于随机森林的女性体型判别
Female body shape prediction based on random forest
尹玲 1夏蕾 2许才国1
作者信息
- 1. 宁波大学 艺术学院,浙江宁波315211
- 2. 上海工程技术大学服装学院,上海201620
- 折叠
摘要
针对目前服装行业亟待解决的服装合体性问题,为准确判断用户体型,分析730名年龄在18 ～ 50岁之间的女性体型数据,提取表征体型特征的6个重要因子.根据特征因子,从整体、局部和躯干轮廓3个层面对女性体型分类.在此基础上,以随机森林理论算法为基础,以R语言为实现工具,建立女性体型分类判别模型.结果显示,3个随机森林分类器的分类精度都较高,训练样本及测试样本的准确率均在85％以上,表明基于随机森林法建立的女性体型判别模型是可靠的.进一步通过随机森林对变量重要性的评估,筛选出表征女性形体指标的重要特征变量.
Abstract
Clothing fit is a problem badly needed prompt solution in the present apparel industries.In order to determine the true shape of female body accurately,the large number of measurement data of 730 female subjects aging 18-50 was analyzed and six characteristic factors were extracted by factors analysis.Female figure was classified from three levels,including the whole body type,local morphological characteristics and torso silhouette.According to it,a prediction model of female body shape was established by using the algorithm of random forests and the programmed tool R language.The results showed that three of the random forest classifiers had high accuracy of prediction,which was up to 85％ both for train samples and test samples.It suggested that the prediction model was reliable for female figure identification.Further,the vital characteristic variables featuring female body shape were filtered by using random forest variable importance measures.
关键词
女性体型/体型分类/判别模型/随机森林/特征变量Key words
female body shape/body shape classification/prediction model/random forest/characteristic variables引用本文复制引用
基金项目
浙江省教育厅科研项目(Y201224639)
出版年
2014
NETL
NSTL
维普
万方数据