摘要
针对区域经济社会系统指标数据存在的高度非线性和二次规划过程涉及的计算复杂性,建立了基于线性规划支持向量回归的区域经济发展预测算法LPSVR。针对预测结果的可信度问题,在LPSVR算法中引入了预测信任度的概念,并给出了具体的计算方法。LPSVR算法在优化过程中折中考虑了经验风险和置信范围,因而具有良好的泛化能力。LPSVR采用线性规划取代二次规划,使得计算的时间和空间复杂度大为降低。以江门市的经济发展预测作为研究实例,LPSVR算法给出了满意的预测结果和预测信任度。
To solve the problems of high nonlinearity in indicator data of regional social systems and computation complexity of quadratic programming, a regional economic development forecasting algorithm based on LPSVR is proposed. The concept and detailed computation methods of forecasting trusting degree is introduced in LPSVR to solve the problem of reliability of forecast data. The introduction of a trade-off between empirical risk and confidence interval in the optimization of LPSVR leads to good generalization performance. LPSVR decreased computational complexity of time and space by replacing quadratic programming with linear programming. An experiment of forecasting economic development of Jiangmen was conducted, and the results and forecasting trusting degree are saris actory.
出处
《辽宁工程技术大学学报(自然科学版)》
EI
CAS
北大核心
2007年第1期129-131,共3页
Journal of Liaoning Technical University (Natural Science)
基金
国家自然科学基金资助项目(70471074)
广东省科技计划基金资助项目(2004B36001051)
中国博士后科学基金资助项目(2005038042)
关键词
支持向量回归
线性规划
预测
区域经济
support vector regression
linear programming
forecasting
regional economy