摘要
为了克服神经网络存在的收敛速度慢、容易陷入局部极值等缺点,提出基于粒子群优化支持向量机(PSO-SVM)的黄金价格预测方法,以影响黄金价格的美元走势、世界黄金储备、石油价格等因素为输入,黄金价格为输出.用粒子群优化算法选择合适的支持向量机参数,对支持向量回归机进行训练.应用训练完成的支持向量回归机预测下一年的黄金价格.结果证明,PSO-SVM的预测精度高于BP神经网络,PSO-SVM适用于黄金价格预测.
In order to overcome the defect of neural network such as slow convergence rate and tendency to fall into local minimum, a gold price forecasting method was presented that based on p~ ticle swarm op- timization with support vector machine （PSO-SVM）, where the dollar trend, world gold reserves, oil price and other factors that have influence on gold price were taken as input, and the gold price was taken as output. The particle swarm optimization algorithm was employed to select the appropriate parameters of support vector machine and the support vector regressive machine was trained. Then, the trained sup- port vector regressive machine was used to forecast the gold price in next year. The result showed that the forecasting accuracy with PSO-SVM was higher than that with BP neural network and PSO-SVM was ap- plicable for gold price forecast.
出处
《兰州理工大学学报》
CAS
北大核心
2013年第3期65-69,共5页
Journal of Lanzhou University of Technology
基金
宁夏自然科学基金(NZ12228)
宁夏高等学校科学研究项目(NJ201279
NJ201233681)
宁夏师范学院创新团队资助项目(ZY201212)
宁夏师范学院重点项目(ZD201311)的资助
关键词
粒子群算法
支持向量机回归
黄金价格
参数优化
统计学习理论
particle swarm algorithm
support vector machine regression
gold price
parameter optimi-zation
statistical learning theory
