【摘要】 在中国棉花目标价格改革的背景下，采用理论分析和自回归分布滞后模型(ARDL)模型研究棉花期货价格的影响因素，并用优化的长短期记忆网络(LSTM)模型对其进行预测，达到较优的预测效果，为揭示棉花期货价格波动特征和主要影响因素提供了理论支持，同时为棉纺织企业把握棉花市场动态、降低棉花成本风险提供了可靠依据。理论分析和ARDL模型结果表明：中国棉花期货价格具有涨跌特征明显、波动幅度大、季节性特征不显著和较强金融属性的特征；中国棉花期货价格主要受自身滞后1期、美棉期货价格、大宗商品价格指数、涤纶短纤价格、黏胶短纤价格、棉花期货成交量等因素的影响；大宗商品价格指数对棉花期货价格影响最大，两者具有显著的联动效应，美棉期货价格影响次之，随后是涤纶短纤和黏胶短纤价格的影响，而自身滞后1期的棉花期货价格和棉花期货成交量的影响较为有限。进一步对棉花期货价格进行预测，结果显示该优化的LSTM模型平均预测误差仅为1.89%,具有良好的预测性能。最后提出了相关政策建议。更多还原

【Abstract】 In the context of cotton target price reform in China, the theoretical analysis and ARDL model are used to research the influencing factors of cotton futures price, and an optimized LSTM model is used to predict it, which yields a favorable prediction result. These findings provide theoretical support for understanding the volatility characteristics and the main influencing factors of cotton futures price, and offer reliable basis for cotton textile enterprises to grasp the cotton market dynamics and to reduce the risk of cotton cost. The theoretical analysis and the ARDL model results show that the cotton futures price exhibits clear fluctuation patterns with large amplitudes, insignificant seasonality and strong financial attributes, which is mainly affected by its own one-period lagged price, the US cotton futures price, the commodity price index, polyester staple fiber price, viscose staple fiber price, and the cotton futures trading volumes. Among these factors, the commodity price index has the greatest impact, displaying a significant linkage effect with cotton futures price. The US cotton futures price has the second-largest impact, followed by polyester staple fiber price and viscose staple fiber price. The impacts of its own one-period lagged price and trading volumes are comparatively minor. Further prediction of cotton futures price reveals that the LSTM model has a good prediction performance with a mean absolute percentage error of only 1.89%. Finally, relevant policy suggestions are provided.更多还原