【摘要】：隨著經(jīng)濟(jì)全球化的發(fā)展,國(guó)際期貨市場(chǎng)中各大類大宗商品價(jià)格波動(dòng)劇烈,而全球經(jīng)濟(jì)形勢(shì)不明朗以及貨幣政策不確定使得大宗商品期貨價(jià)格難以被準(zhǔn)確預(yù)測(cè).本文選取玉米,原油,黃金分別作為大宗商品農(nóng)產(chǎn)品類、能源類、金屬類的代表對(duì)象,基于奇異譜分析方法(singular spectrum analysis,SSA),對(duì)商品期貨價(jià)格進(jìn)行分解,結(jié)合Kmeans動(dòng)態(tài)聚類技術(shù)將分解量聚合成不同特征的價(jià)格序列,再采用具有優(yōu)良特性的極限學(xué)習(xí)機(jī)算法(extreme learning machine,ELM)對(duì)模型進(jìn)行訓(xùn)練,得到大宗商品期貨價(jià)格預(yù)測(cè)模型.實(shí)證結(jié)果表明,采用序列分解聚類策略能夠顯著提高模型預(yù)測(cè)精度,在價(jià)格未來(lái)的整體水平和變動(dòng)方向上都能達(dá)到較好的預(yù)測(cè)效果.
[Abstract]:With the development of economic globalization, commodity prices in international futures markets fluctuate sharply, while global economic uncertainty and monetary policy uncertainty make it difficult to accurately predict commodity futures prices. In this paper, corn, crude oil and gold are selected as the representative objects of commodity agricultural products, energy and metals, and the commodity futures prices are decomposed based on the singular spectrum analysis method (singular spectrum analysis,SSA). Combined with Kmeans dynamic clustering technology, the decomposed quantities are aggregated into price sequences with different characteristics, and then the model is trained by the extreme learning machine (extreme learning machine,ELM) algorithm with excellent characteristics, and the commodity futures price prediction model is obtained. The empirical results show that the prediction accuracy of the model can be improved significantly by using the sequence decomposition and clustering strategy, and the prediction results can be achieved in the overall level and the direction of change of the price in the future.
【作者單位】： 中國(guó)科學(xué)院數(shù)學(xué)與系統(tǒng)科學(xué)研究院;中國(guó)科學(xué)院國(guó)家數(shù)學(xué)與交叉科學(xué)中心;中國(guó)科學(xué)院大學(xué);北京科技大學(xué)數(shù)理學(xué)院;
【基金】：國(guó)家自然科學(xué)基金(71271202) 中國(guó)科學(xué)院青年創(chuàng)新促進(jìn)會(huì)項(xiàng)目~~
【分類號(hào)】：F713.35;TP18
，

本文編號(hào)：2203369