Research On Prediction Model Of Repoverty Based On Logistic Regression Analysis

Targeted poverty alleviation is an important strategic measure to build a well-off society in our country.Though remarkable achievements of poverty alleviation,there is a strange circle phenomenon of poverty alleviation,elimination and repoverty in many poor areas for a long time.In response to the call of the National Targeted Poverty reduction Strategy,this thesis studies the phenomenon of repoverty in China by using big data analytics.In the data analysis module of the first management platform of targeted poverty alleviation in Gansu Province,this study designs prediction model of repoverty based on Logistic regression analysis.Firstly of all,the preprocessing work of the missing values,outliers,data transformation provided by Gansu Poverty Alleviation Office is carried out,then the independent sample T test is made on each characteristic of rural poverty population,the correlation test is performed on the characteristics whose significance level is lower than 0.05 and feature modeling is selected.Secondly,based on the Spark platform,the Logistic regression algorithm is used to analyze the existing poverty's return.According to the experimental results,the model was evaluated and the optimal model with an accuracy rate of 86.43% was selected to predict repoverty in 2017.Finally,the K-means clustering algorithm is used to analyze the predicted returning poor population,and the optimal K value is selected to set up the cluster algorithm.The differences in the characteristics of all kinds of returning poor people are observed through the clustering results.This research will analyze the reasons why all kinds of returning poor households may return to poverty,set up corresponding investigation groups for different categories of people returning to poverty,formulate effective measures and help the poor get rid of the bad luck of returning to poverty.