在我国金融经济体系中,农产品期货市场不仅能引导市场自我调节,还为监管者提供了高效的信息传递渠道。有效、准确地预测期货价格有助于指导农业生产、监控价格波动带来的经营风险,并提升宏观调控政策的预见性与精准性。本文主要探讨粮...在我国金融经济体系中,农产品期货市场不仅能引导市场自我调节,还为监管者提供了高效的信息传递渠道。有效、准确地预测期货价格有助于指导农业生产、监控价格波动带来的经营风险,并提升宏观调控政策的预见性与精准性。本文主要探讨粮食期货市场中的玉米期货价格预测问题,研究以大连商品交易所2005至2023年期间的玉米连续期货日度基本指标和技术指标数据为样本,构建了一种基于多层感知器(MLP)的玉米期货价格预测模型,用于拟合玉米期货的收盘价。考虑到预测精度与计算效率的平衡,本文还引入了遗传算法(GA)对MLP模型的参数进行优化,以提高预测结果的准确性。此外,为了全面评估所提出模型的表现,本文将优化后的MLP模型与决策树(DT)、随机森林(RF)、XGBoost和LightGBM等多种主流机器学习算法进行了对比分析。实证结果表明,MLP模型在MSE、MAE、R2等多个评估指标上优于其他四个基准模型,表现出更强的预测能力。而通过遗传算法(GA)对MLP模型参数进行优化后,模型的预测性能得到了进一步提升,尤其在价格波动较大的市场环境下,优化后的MLP模型展现了较好的稳健性和精确性。In China’s financial and economic system, the agricultural futures market not only guides market self-regulation, but also provides efficient information transmission channels for regulators. Effectively and accurately predicting futures prices helps guide agricultural production, monitor operational risks caused by price fluctuations, and enhance the predictability and precision of macroeconomic regulation policies. This article mainly explores the problem of predicting corn futures prices in the grain futures market. Using the daily basic and technical indicators of corn futures from 2005 to 2023 on the Dalian Commodity Exchange as samples, a corn futures price prediction model based on multi-layer perceptron (MLP) is constructed to fit the closing price of corn futures. Considering the balance between prediction accuracy and computational efficiency, this article also introduces genetic algorithm (GA) to optimize the parameters of the MLP model to improve the accuracy of prediction results. In addition, in order to comprehensively evaluate the performance of the proposed model, this paper compared and analyzed the optimized MLP model with various mainstream machine learning algorithms such as decision tree (DT), random forest (RF), XGBoost, and LightGBM. The empirical results show that the MLP model outperforms the other four benchmark models in multiple evaluation indicators such as MSE, MAE, and R2, demonstrating stronger predictive ability. After optimizing the parameters of the MLP model through genetic algorithm (GA), the predictive performance of the model was further improved, especially in market environments with significant price fluctuations. The optimized MLP model demonstrated good robustness and accuracy.展开更多
本文运用长短期记忆网络(LSTM)和变换器(Transformer )混合模型预测豆油期货价格。通过数据处理、模型构建、训练评估与结果分析,发现该模型能有效地捕捉价格序列特征,在测试集上展现出良好预测性能,在平均绝对误差(MAE),均方根误差(RMS...本文运用长短期记忆网络(LSTM)和变换器(Transformer )混合模型预测豆油期货价格。通过数据处理、模型构建、训练评估与结果分析,发现该模型能有效地捕捉价格序列特征,在测试集上展现出良好预测性能,在平均绝对误差(MAE),均方根误差(RMSE),平均绝对百分比误差(MAPE),相对均方根误差(RRMSE)等指标上表现优异,与单独的Transformer模型和LSTM模型预测相比,精确度有明显的提高。这表示该模型在期货价格预测领域具有一定的应用潜力。This paper uses a hybrid model of Long Short-Term Memory (LSTM) and Transformer to predict the futures price of soybean oil. Through data processing, model construction, training evaluation, and result analysis, it is found that this model can effectively capture the characteristics of the price sequence. It demonstrates excellent prediction performance on the test set and shows outstanding performance in indicators such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Relative Root Mean Square Error (RRMSE). Compared with the predictions of the individual LSTM model and Transformer model, the accuracy has been significantly improved. This indicates that this model has certain application potential in the field of futures price prediction.展开更多
文摘在我国金融经济体系中,农产品期货市场不仅能引导市场自我调节,还为监管者提供了高效的信息传递渠道。有效、准确地预测期货价格有助于指导农业生产、监控价格波动带来的经营风险,并提升宏观调控政策的预见性与精准性。本文主要探讨粮食期货市场中的玉米期货价格预测问题,研究以大连商品交易所2005至2023年期间的玉米连续期货日度基本指标和技术指标数据为样本,构建了一种基于多层感知器(MLP)的玉米期货价格预测模型,用于拟合玉米期货的收盘价。考虑到预测精度与计算效率的平衡,本文还引入了遗传算法(GA)对MLP模型的参数进行优化,以提高预测结果的准确性。此外,为了全面评估所提出模型的表现,本文将优化后的MLP模型与决策树(DT)、随机森林(RF)、XGBoost和LightGBM等多种主流机器学习算法进行了对比分析。实证结果表明,MLP模型在MSE、MAE、R2等多个评估指标上优于其他四个基准模型,表现出更强的预测能力。而通过遗传算法(GA)对MLP模型参数进行优化后,模型的预测性能得到了进一步提升,尤其在价格波动较大的市场环境下,优化后的MLP模型展现了较好的稳健性和精确性。In China’s financial and economic system, the agricultural futures market not only guides market self-regulation, but also provides efficient information transmission channels for regulators. Effectively and accurately predicting futures prices helps guide agricultural production, monitor operational risks caused by price fluctuations, and enhance the predictability and precision of macroeconomic regulation policies. This article mainly explores the problem of predicting corn futures prices in the grain futures market. Using the daily basic and technical indicators of corn futures from 2005 to 2023 on the Dalian Commodity Exchange as samples, a corn futures price prediction model based on multi-layer perceptron (MLP) is constructed to fit the closing price of corn futures. Considering the balance between prediction accuracy and computational efficiency, this article also introduces genetic algorithm (GA) to optimize the parameters of the MLP model to improve the accuracy of prediction results. In addition, in order to comprehensively evaluate the performance of the proposed model, this paper compared and analyzed the optimized MLP model with various mainstream machine learning algorithms such as decision tree (DT), random forest (RF), XGBoost, and LightGBM. The empirical results show that the MLP model outperforms the other four benchmark models in multiple evaluation indicators such as MSE, MAE, and R2, demonstrating stronger predictive ability. After optimizing the parameters of the MLP model through genetic algorithm (GA), the predictive performance of the model was further improved, especially in market environments with significant price fluctuations. The optimized MLP model demonstrated good robustness and accuracy.
文摘本文运用长短期记忆网络(LSTM)和变换器(Transformer )混合模型预测豆油期货价格。通过数据处理、模型构建、训练评估与结果分析,发现该模型能有效地捕捉价格序列特征,在测试集上展现出良好预测性能,在平均绝对误差(MAE),均方根误差(RMSE),平均绝对百分比误差(MAPE),相对均方根误差(RRMSE)等指标上表现优异,与单独的Transformer模型和LSTM模型预测相比,精确度有明显的提高。这表示该模型在期货价格预测领域具有一定的应用潜力。This paper uses a hybrid model of Long Short-Term Memory (LSTM) and Transformer to predict the futures price of soybean oil. Through data processing, model construction, training evaluation, and result analysis, it is found that this model can effectively capture the characteristics of the price sequence. It demonstrates excellent prediction performance on the test set and shows outstanding performance in indicators such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Relative Root Mean Square Error (RRMSE). Compared with the predictions of the individual LSTM model and Transformer model, the accuracy has been significantly improved. This indicates that this model has certain application potential in the field of futures price prediction.
