基于随机森林算法的贫困精准识别模型研究

Research on Targeted Poverty Identification Model Based on Random Forest Algorithms

扶贫对象的精准识别是实现精准扶贫的重要条件。实现贫困数据的精准分类与识别以及贫困识别由定性到定量、由单维瞄准向多维瞄准的转变是精准扶贫的重要基础。精准识别可以采用大数据分析中的分类算法实现。本文基于可持续生计分析框架,从人力资本、社会资本、自然资本、物质资本、金融资本和生计环境六个方面建立了多维贫困指标体系,运用随机森林算法构建了精准识别模型,并采用中国家庭追踪调查数据(CFPS),对扶贫对象精准识别模型的分类及识别效果进行了评价,结果表明模型效果良好。

Targeted identification of poverty alleviation objects is an important precondition to achieve targeted poverty alleviation.Realization of classification and identification of big data,conversion from quantity analysis to quality analysis and change from linear targeting to multi-dimensional targeting are the basis of targeted poverty alleviation.Targeted identification can be achieved by using classification algorithm in big data analysis.Based on the framework of sustainable livelihood analysis,this paper establishes a multi-dimensional poverty index system based on sustainable livelihood from six aspects:human capital,social capital,natural capital,physical capital,financial capital and living environment.Using stochastic forest algorithm to construct a targeted identification model and the data done by Institute of Social Science Survey,Peking University,this paper evaluates the effect of the model for classification and identification of poverty alleviation objects.The results show its validity and reliability.