基于动态模型平均的大豆期货价格预测研究

针对大豆期货价格波动的复杂性及影响因素的多元性,本文将动态模型平均理论引入大豆期货价格分析与预测研究中,通过动态选择解释变量和系数时变程度,在有效控制模型和系数不确定性的同时,最大限度综合利用大豆期货市场内外部信息,以提高大豆期货价格预测准确度。具体的,本文提出一套基于动态模型平均理论的大豆期货价格影响因素与预测分析框架,从期货市场和经济环境等两方面准确地识别出大豆期货价格影响因素的时变特征,进而构建大豆期货价格预测模型,并通过预测误差指标和Diebold-Mariano检验法评估其与基准模型的预测能力。研究结果表明,动态模型平均理论在有效剖析大豆期货价格影响因素的时变特征的同时,能明显提升大豆期货价格预测准确度。

Soybean Future Prices Forecasting based on Dynamic Model Averaging

In view of the complexity of soybean futures price fluctuation and the diversity of influencing factors, soybean futures price forecasting is conducted by introducing the dynamic model averaging theory. It should be noted that this technique is capable of dynamically choosing the explanatory variables and coefficient of variation, which will maximize the utilization of various information to control the models effectively and coefficients uncertainty, and finally improve the forecasting performance. More specifically, an analysis framework on influencing factors of soybean futures price is proposed. The time-vary characteristics of soybean futures price's influencing factors are identified from the perspective of futures markets and economics environment, and then a forecasting model for soybean futures price is constructed. Furthermore, covering an out-of-sample period from July 30, 2009 to June 15, 2017, the forecasting performance of the proposed forecasting model is evaluated and compared with six benchmarks on the basis of accuracy measures and Diebold-Mariano test. The experimental results show that the dynamic model averaging can effectively identify the influence degree of each explanatory variables on soybean futures price, and at the same time, outperform the Bayes model averaging, time-varying parameter model, and random walk in soybean futures price forecasting. Policymakers should be cognizant of the fact that there are many potential predictors that can help to forecast the Chinese soybean futures price, and the predictive powers of these predictors vary over time.