Research On Poverty Identification Of Urban Households Under The Background Of Precision Poverty Alleviation

Poverty is one of the most acute social problems in the world today,and it has always received great attention from governments.With the rapid progress of the rural poverty alleviation strategy,the problem of urban poverty has gradually entered the public's field of view and has attracted everyone's attention.In March 2018,Jiangxi Province took the lead in promulgating the "Opinions on Enhancing the Poverty Relief and Poverty Alleviation of Urban Poor People" in the country,explicitly including urban poor people in poverty alleviation and poverty alleviation work,and helping urban poor people to alleviate poverty.The Nineteenth National Congress of the Communist Party of China pointed out: "From now to 2020,it is a period of victory in building a well-off society in an all-round way." If the problem of urban poverty is not resolved,this problem will become a stumbling block in the development of a comprehensive well-off society.Therefore,this paper uses the CFPS2018 database of the latest round of survey data from the Chinese Family Tracking Survey(CFPS)to study the identification of poverty in urban households.First,based on the existing research on poverty identification in rural households,this paper refers to the Rural Multidimensional Poverty Index(MPI),combines the current urban poverty situation in China and the characteristics of urban families,and finally starts with demographic characteristics,health,education,basic living security,and living conditions.Five dimensions build the urban family poverty identification index system.The demographic dimension includes three indicators: family size,age structure,and labor force ratio.The health dimension selects the health index and the education dimension selects education.As an indicator of degree,basic living security includes three indicators: drinking water,electricity,and fuel.The living conditions dimension includes seven indicators: per capita annual income,personal income tax,Engel coefficient,durable consumer goods,per capita housing area,living security,and medical security.After constructing the index system,the independent variables and dependent variables were determined based on the contents of the CFPS database,and the correlations between the variables and urban household poverty were verified through contingency table chi-square test and independent sample t test.Secondly,the pre-processed sample data is divided into a 3: 7 test set and a training set,and the training set is used to establish models such as logistic regression,classic decision trees,random forests,and support vector machines to identify urban household poverty;Set to verify the classification prediction capabilities of these models.The accuracy of the final four models is about 94%,indicating that the index system constructed in this paper is scientific;in addition,the prediction accuracy of logistic regression and random forest models is better than that of classic decision trees and support vector machine models..Finally,by comparing the accuracy rate,positive hit rate,negative hit rate,recall rate,specificity,F1 value and other indicators of the classification model evaluation,we conclude that logistic regression and random forest model are more effective To be effective,the prediction accuracy and stability are better than other models.