R软件3个常用决策树包的应用比较

The Comparison of the Three Decision Tree Packages in R Software

目的通过应用R软件中决策树包rpart、party、partykit对伤害发生的影响因素进行分析，比较3种方法分析结果的差别，为伤害资料分析方法提供一种新手段。方法根据农村女性伤害发生专题调查资料，在R软件中分别建立决策树包rpart、party、partykit的模型，比较3种方法分析结果的差异。结果rpart包与party、partykit包分类规则不一致。rpart包分类结果表明未接受教育的农民女性是伤害的高危人群，party包与partykit包分类结果表明接受教育女性中离婚或丧偶的女性伤害发生率最高，其次是未接受教育的女性。结论R软件决策树包的应用条件有所不同，应该根据实际情况选择合适的软件包进行应用分析。在伤害资料的分类建模分析中，使用partykit包较为适合、简捷且界面友好。

Objective The three Decision Tree packages in R software of rpart,party and partykit were compared in terms of those factors affecting the occurrence of injuries so as to provide a new way in the analysis of injury data.Methods The rpart,party,and partykit decision tree models were established based on the data of a previous study,and comparisons were made with the results produced by the three models.Results The classification rule of rpart package was different from that of party and partkit package.rpart package classification results showed that the rural women with lower education levels is a high-risk group.The divorced or widowed women with higher education levels had the highest incidence of injury compared to the women with lower education levels by party and partykit package.Conclusions Cautions should be exercised when selecting an appropriate package in practice;for the classification of injury data and modeling of them,the partykit package is more suitable because of its friendly and beautiful interfaces.