基于奇異譜分解的玉米價格多尺度組合預測

張大斌 陳政希 黃玉宣 王楠 林婉 潘鎮飛

摘要為提高玉米價格預測精度，基于分解-重構-集成思想，構建一個基于奇異譜分解的多尺度組合模型。首先對原始序列進行奇異譜分解，并用對角平均法將分量序列重構，用作單個模型的輸入，最后用BP神經網絡對各單一模型輸出進行非線性集成。對比分析了分別將原始序列、重構序列作為輸入，各單一模型和多尺度組合模型的預測效果。結果表明，該研究所建模型要優于各單一模型。基于價格總體趨于平穩的情況，政府應結合實際情況適當采取措施以保障玉米價格的持續穩定。

關鍵詞玉米價格;奇異譜分解;多尺度模型;組合預測

中圖分類號S-9文獻標識碼A

文章編號0517-6611（2020）15-0241-06

doi：10.3969/j.issn.0517-6611.2020.15.068

開放科學（資源服務）標識碼（OSID）：

Multiscale Combination Forecast of Corn Price Based on Singular Spectrum Decomposition

ZHANG Dabin，CHEN Zhengxi，HUANG Yuxuan et al

（School of Mathematics and Information （School of Software）， South China Agricultural University，Guangzhou，Guangdong 510642）

AbstractIn order to improve the accuracy of corn price forecasting， a multiscale combination model based on singular spectral decomposition is constructed based on the decompositionreconstructionintegration idea.First， singular spectral decomposition is performed on the original sequence， and the component sequence is reconstructed using the diagonal average method as the input of a single model. Finally， the BP neural network is used to integrate the outputs of each single model nonlinearly. The prediction effect of each single model and multiscale combination model was compared and analyzed using the original sequence and the reconstructed sequence as inputs.The results showed that the model built in this paper is better than each single model.Based on the fact that prices are generally stable， the government should take appropriate measures in accordance with actual conditions to ensure the continued stability of corn prices.

Key wordsCorn price;Singular spectrum decomposition;Multiscale model;Combined forecast

基金項目廣東省自然科學基金項目（2016A030313402）;大學生創新創業項目（201810564106）。

作者簡介張大斌（1969—），男，湖北潛江人，教授，博士生導師，從事預測理論與方法研究。

收稿日期2019-12-03;修回日期2020-01-09