摘要
随着经济全球化的发展,国际期货市场中各大类大宗商品价格波动剧烈,而全球经济形势不明朗以及货币政策不确定使得大宗商品期货价格难以被准确预测.本文选取玉米,原油,黄金分别作为大宗商品农产品类、能源类、金属类的代表对象,基于奇异谱分析方法(singular spectrum analysis,SSA),对商品期货价格进行分解,结合Kmeans动态聚类技术将分解量聚合成不同特征的价格序列,再采用具有优良特性的极限学习机算法(extreme learning machine,ELM)对模型进行训练,得到大宗商品期货价格预测模型.实证结果表明,采用序列分解聚类策略能够显著提高模型预测精度,在价格未来的整体水平和变动方向上都能达到较好的预测效果.
With the development of economic globalization, the fluctuations of commodity futures prices are increasingly violent. The uncertainty of the global economic situation and monetary policy make the commodity futures prices difficult to be accurately predicted. In this paper, corn, gold and crude oil are selected as the representatives of agricultural futures, industrial metals futures and energy futures, respectively. Based on the singular spectrum analysis （SSA） method, we decompose the commodity futures price and incorporate Kmeans dynamic clustering technique to group the decomposition series. Then the forecasting models of commodity futures price are developed by using extreme learning machine （ELM）. The empirical results show that the decomposition and clustering schemes can significantly improve the accuracy of price forecasting, and performs well on the overall estimation and direction of change of the commodity prices.
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2017年第8期2004-2014,共11页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(71271202)
中国科学院青年创新促进会项目~~
