Research On The Performance Of Targeted Poverty Alleviation Based On RS-SVM Model

In the current context of targeted poverty alleviation,how to scientifically and effectively evaluate the performance of poverty alleviation is a question worth exploring.On the one hand,it can comprehensively evaluate the work effect,on the other hand,it plays a promoting role in the sustainable development of poverty-stricken counties.Through literature and the present research results,this paper found that the current scholars for the performance study of poverty alleviation,there has been a relatively abundant achievements,there are many of the spirit is worth our using for reference,but there are also some shortcomings,such as the evaluation method of choice,the commonly used analytic hierarchy process(ahp),etc.,often based on the experience of expert assignment,subjectivity is stronger.Factor analysis method is relatively objective,but if the data does not pass the KMO test,the factor analysis shows inapplicability,and the actual meaning of the comprehensive factor is difficult to precisely define.In this paper,a performance evaluation method for poverty alleviation is proposed,which takes rough set as the front system of support vector machine modeling and applies the two to the performance evaluation of targeted poverty alleviation.On the basis of theoretical research,the advantages and complementarities of the two are analyzed,and the construction steps of the model are explained in detail.Based on the empirical analysis of 23 poverty-stricken counties in Shanxi Province,the index system was constructed with 14 indicators from four levels: economic benefit,social benefit,ecological benefit and anti-poverty.Then,the rough set theory is used to reduce the index attributes of the initial decision table.The importance of each attribute is calculated based on the attribute importance.The unimportant indexes are reduced and the corresponding weights are obtained.Then support vector machine(SVM)model is established for the new decision table,in the process of parameter optimization,using the grid searchmethod,genetic algorithm and particle swarm optimization(pso)algorithm,cross-validation on the training set,respectively,to establish forecasting model,the experimental results show that three kinds of algorithms can effectively optimize the support vector machine parameters,model show the excellent generalization performance.Compared with the traditional support vector machine modeling,the results show that the rs-svm model improves the prediction accuracy.In this study,rs-ga-svm was selected as the optimal model based on the calculation time of the model,the cross-validation error rate and the predicted mean square error.By comparing the factor analysis method with the prediction and evaluation results in this paper,it can be seen from the results that the prediction results of the model used in this paper are closer to the reality and the universal applicability of the model is wider.The advantages and feasibility of the model have been verified in the practical application,which makes up for the subjectivity and low universality of typical performance evaluation methods,enriches the performance evaluation method of targeted poverty alleviation,and has certain practical reference significance in the performance research of targeted poverty alleviation.