基于偏最小二乘回归的城镇登记失业率预测  

Urban Registered Unemployment Rate Forecasting Base on Partial Least-Squares Regression

作　　者：韩庆艳[1] 李文涛[2] 

机构地区：[1]安徽邮电职业技术学院,合肥230031 [2]陆军军官学院,合肥230031

出　　处：《舰船电子工程》2013年第12期100-103,共4页Ship Electronic Engineering

摘　　要：选取了城镇登记失业率的五类影响因素和18个预测指标,采用本征值与病态指数检验法进行多重共线性进行诊断,说明在18个预测指标之间存在着严重的多重共线性,为了克服多重共线性导致的预测失真问题,运用偏最小二乘回归模型,以我国城镇登记失业率为研究对象,按照交叉有效性的决策原则,建立了预测模型,所提取成分对X累计解释能力达到99.6%,对y的累计解释能力达到96%,具有较高的预测精度,该方法还可进一步应用到类似问题中。Firstly, the urban registered unemployment rate in the 5 kind of influence factors and the 18 predictors are determined, then the ei- genvalue and the morbid state index inspection method are used to carry on the diagnosis of multicollinearity, the results show that exists severe multi- collinearity between 18 predictors, Secondly in order to overcome the prediction distortion caused by multicollinearity problems, the partial least-squares model is applied to forecasting the urban registered unemployment rate of China, the model has a higher forecast accuracy, its cumulative interpretation ability of X and y are 99.60/00 and 96~ , the method can also be further applied to similar problems,

关 键 词：多重共线性 偏最小二乘回归 交叉有效性 解释能力 城镇登记失业率 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]