基于支持向量机的养老保障满意度非线性模型

Nonlinear Model for Satisfaction of Old-Age Security Based on Support Vector Machine

目前对养老保障满意度的研究所采用的统计方法,都是基于养老保障满意度与影响因素之间的线性关系,且未对所建模型进行检验及理论预测。由于事物之间关系复杂,变量之间往往呈现非线性关系。采用支持向量机算法结合粒子群优化算法,建立养老保障满意度非线性模型。用于研究的养老保障满意度样本数为8 339份。结果显示,基于支持向量机的分类模型对养老保障满意度预测精度高于76%,预测性能优于二元逻辑回归预测结果。表明养老保障满意度与受教育程度、受教育满意度、家庭经济状况满意度、总体生活满意度、对社会总体评价等5个影响因素之间存在非线性关系。因此,应用支持向量机算法建立养老保障满意度非线性模型是可行的。

Currently,the statistical methods used for the satisfaction of the old-age security are based on the assumption that there are linear relationships between the satisfaction of the old-age security and the influencing factors. Moreover,these models have not been tested and predicted theoretically. In fact,these relationships should be nonlinear due to the complexity of the relationships. This paper is the first report on the development of the nonlinear model for satisfaction ofthe old-age security,by applying support vector machine(SVM),together with particle swarm optimization(PSO). The number of samplesfor the satisfaction of oldage security is 8339. Results show that the accuracy of SVM classification in predicting satisfaction of old-age security is above 76%,and the prediction performance based on SVM classification is better than that of binary logistic regression(BLR). Results indicate that there are non-linear relationships between the satisfaction of the old-age security and the five influencing factors(educational level,educational satisfaction,family economic status satisfaction,overall life satisfaction,and overall social evaluation).The feasibility of applying PSOSVM to build nonlinear model for satisfaction of old-age security has been demonstrated.