基于核主成分与加权支持向量机的福建省城镇登记失业率预测

Forecasting of the urban registered unemployment rate in Fujian province based on kernel principal component analysis and weighted support vector machine

选取了影响失业率的19个指标,构建了基于核主成分分析与加权支持向量机的预测方法,给出了具体的预测步骤,并用此方法对福建省城镇登记失业率进行了预测研究.研究结果表明,由于所用预测方法考虑了指标的相关性及不同时期样本的不同重要性并进行了简化降维,拟合及预测都达到了很高的精度,其相对误差都小于1%,说明用核主成分分析与加权支持向量机来预测失业率是可行且有效的,并可将其推广到其它领域的预测问题.

The key of the early warning system of unemployment is the unemployment rate forecasting. The high precision forecasting to the unemployment rate could offer lots of valuable information for the government policies making. In this paper, 19 indices that affect the unemployment rate were chosen, and a new forecasting method based on the kernel principal component analysis and the weighted support vector machine was proposed. Further more, the detailed forecasting steps were also given, and then the new method was applied to the urban registered unemployment rate forecasting in Fujian province, The research results indicate that the precision of fitting and forecasting are very high, and all the relative errors are less than 1%. This is because the new forecasting method considers the correlativity between the indices and the different importance of the samples in different periods, and simplifies the index system by dimension reduction. The results show that, the new method on the basis of the kernel principal component analysis and the weighted support vector machine is feasible and valid to forecast the unemployment rate, and it also can be generalized to the forecasting of other domain.