Research On Identification Of Poor Households In Precision Poverty Alleviation In Jiangxi Province

Since 1986,China has continuously promoted poverty alleviation and development work.After years of hard work,it has made tremendous achievements in poverty alleviation.In order to solve the poverty problem in the new era in a deeper way,the Fifth Plenary Session of the 18 th Central Committee in 2013 put forward the idea of “precise poverty alleviation” and issued a series of policies to promote poverty alleviation.The provinces and cities across the country also actively responded to the call of the state and committed to complete poverty alleviation work.The old revolutionary province of Jiangxi Province is one of the major rural population provinces in China and the main battlefield for poverty alleviation and development in China.In order to ensure the implementation of poverty alleviation work,the Jiangxi Provincial Government has implemented corresponding measures to implement policies to the village to the household.However,in practice,it has been found that it is difficult to obtain reliable income and expenditure information from the household level,and the implementation of poverty investment projects has long relied on the targeting of regional targets.The resulting incomplete coverage and leakage problems require a simpler and more effective means of poverty targeting to identify poor families.At present,some scholars have applied the identification model to the process of identifying poor households.However,most of the poor household identification models in China still use Logistic regression to establish models.However,this regression model based on hypothesis is prone to errors.This leads to inaccurate recognition.In this regard,based on the previous studies,this paper makes a further discussion on the identification model of poor households in Jiangxi Province.In theory,it introduces the current research on precision poverty alleviation and related research on applying machine learning algorithms to various fields,and discusses some advantages of random forest and support vector machine in classification.These studies will be machine learning later.The method used to pave the way for the identification of poor households.In terms of indicators,combined with the actual situation in Jiangxi Province,the existing indicator system has been expanded to better reflect the poverty situation of farmers.In terms of method,this paper applies the machine learning method to the problem of poverty recognition,and uses the learning curve,the Brier score and the probability calibration to measure the validity and stability of several models.The empirical analysis shows that the support vector machine model is superior to the logistic model and the random forest model,and the test set data is used to prove the stability of the model's prediction effect,which can bring certain reference value to the poor household identification work.The thesis is divided into six chapters.The first chapter introduces the background and significance of the paper,the related research status,and then introduces the research ideas,methods and contents,and finally points out the innovation of this paper.The second chapter introduces the relevant theory of the model,including the basic idea of the model and the measure of the performance of the classifier.The third chapter mainly selects poverty indicators and descriptive statistics of poverty data.In the fourth chapter,this paper uses the logistic method commonly used by scholars and the newly introduced machine learning method(random forest,support vector machine)to build the model on the training set,which lays a foundation for the comparison between the later models.In the fifth chapter,the model is tested using the data of the test set.The results show that the support vector machine model is superior to the random forest model and the logistic model in terms of accuracy,recall rate and predictive ability,and the classification effect and stability of the model are more ideal.Finally,in the summary,the support vector machine does not identify the correct farmers,find out the possible problems,and proposes not only to pay attention to the income indicators in the future poverty identification,but also to consider the income composition.In the sixth chapter,the conclusions of this paper are summarized,the shortcomings are pointed out and the outlook is put forward,and some policy recommendations are put forward.