基于多维度综合决策的猪肉价格时间序列预测模型

Time Series Prediction Model for Pork Prices Based on Multi-dimensional Comprehensive Decision-making

摘要

山东省是畜牧业大省，猪肉价格波动对居民生活质量有着重大影响，目前，关于山东省生猪价格波动的预测研究较少，且存在预测时间较短，时间窗口狭窄和预测结果不准确等问题。针对传统预测模型存在长时序预测准确率不足的问题，本文提出了一种基于综合决策机制的时间序列预测模型。首先，将时间序列信息进行分解，通过可逆归一化将数据特征进行放大，以提取更多的价格波动信息；在信息分解的基础上，通过上采样扩充先验知识，并采用多维度综合决策的方式，增强多层感知机的数据特征挖掘能力和决策能力；最后，将先验知识与预测结果进行直接映射，解决了窗口狭窄和滑动窗口迭代预测导致的误差累积问题。试验结果表明，相较于ARIMA、Prophet-BP、GA-BP、VMD-LSTM 和 STL-Informer 模型，本文算法在 RMSE（Root mean square error）和 MAE（Mean absolute error）指标上平均提升了50.2% 和30.9%，在 R ²（Coefficient of Determination）指标上的稳定性优于上述对比算法，平均提升了60.2%。本文所提出的算法对于山东省生猪市场的预测性能更优，有助于相关部门对生猪价格波动做出科学决策。

Abstract

Shandong Province is a major hub for animal husbandry, and pork price fluctuations have a significant impact on residents' quality of life. Currently, there is limited research on predicting hog price fluctuations in Shandong Province, and issues such as short prediction horizons, narrow time windows, and inaccurate forecasting results persist. To address the insufficient accuracy of traditional prediction models in long-term time series prediction, this paper proposes a time series prediction model based on a comprehensive decision-making mechanism. Firstly, this study decomposes time series information and amplifies data features through reversible normalization to extract more price fluctuation information. Based on the information decomposition, it expands prior knowledge via upsampling and enhances the data feature mining and decision-making capabilities of the multi-layer perceptron through multi-dimensional comprehensive decision-making. Finally, it directly maps prior knowledge to prediction results, thereby addressing the issues of narrow windows and error accumulation caused by sliding window iterative prediction. The experimental results show that compared with models such as ARIMA, Prophet-BP、 GA-BP, VMD-LSTM, and STL-Informer, the algorithm in this paper achieves an average improvement of 50.2% and 30.9% in RMSE and MAE indicators, respectively. Furthermore, it exhibits superior stability in the R² indicators, with an average improvement of 60.2% over the aforementioned comparative models. The proposed algorithm exhibits better forecasting performance for the hog market in Shandong Province, which can assist relevant departments in making scientific decisions regarding hog price fluctuations.

关键词

猪肉价格 / 时间序列预测 / 趋势分解 / 上采样 / 多维度决策

Key words

Pork price / time series prediction / trend decomposition / up-sampling / multi-dimensional decision

参考文献

[1] 李俊茹, 石自忠, 胡向东. 中国生猪市场价格波动的双疫情冲击效应[J]. 统计与决策, 2023, 39(17): 79-83.

[2] 李霞, 陶建平, 涂涛涛. 突发动物疫情信息与猪肉价格波动——基于消费者网络关注视角[J]. 农业技术经济, 2024(02): 73-95.

[3] 陈怡然, 熊竹青, 周脚根, 等. 畜禽养殖业数据应用展望和问题分析[J]. 中国科学院院刊, 2024, 39(11): 1982-1993.

[4] 石庆兰, 束金阳, 李道亮, 等. 基于BiLSTM-GRU融合网络的稻虾养殖溶解氧含量预测[J]. 农业机械学报, 2023, 54(10): 364-370.

[5] 杨亮, 刘春红, 郭昱辰, 等. 基于EMD-LSTM的猪舍氨气浓度预测研究[J]. 农业机械学报, 2019, 50(S1): 353-360.

[6] 刘春红, 杨亮, 邓河, 等. 基于ARIMA和BP神经网络的猪舍氨气浓度预测[J]. 中国环境科学, 2019, 39(06): 2320-2327.

[7] 王润周, 张新生, 王明虎. 基于信号分解和深度学习的农产品价格预测[J]. 农业工程学报, 2022, 38(24): 256-267.

[8] 任青山, 方逵, 朱幸辉. 基于多元回归的BP 神经网络生猪价格预测模型[J]. 江苏农业科学, 2019, 47(14): 277-281.

[9] 吴培, 李哲敏 . 中国猪肉价格预测研究- 基于ARIMA-GM-RBF 组合模型的分析[J]. 价格理论与实践, 2019(01): 75-78.

[10] 魏嘉妮. 基于GA-BP 神经网络模型的猪肉价格预测分析[D].大连:东北财经大学, 2023.

[11] 刘合兵, 华梦迪, 席磊, 等. 白条猪价格预测模型构建[J]. 河南农业大学学报, 2024, 58(01): 123-131.

[12] 胡春安, 江维. 基于VMD-BO-BiLSTM的猪肉价格预测模型[J]. 应用科学学报, 2023, 41(04): 692-704.

[13] 叶勇, 许潘, 辜丽川, 等. 基于Prophet-BP 模型的猪肉价格预测研究[J]. 黄山学院学报, 2021, 23(05): 75-80.

[14] Guo Y, Tang D, Cai Q, et al. Agricultural price prediction based on data mining and attention-based gated recurrent unit:a case study on China’s hog[J]. Journal of Intelligent & Fuzzy Systems, 2024, 46(4): 9923-9943.

[15] Sun C, Pei M, Cao B, et al. A study on agricultural commodity price prediction model based on secondary decomposition and long short-term memory network[J]. Agriculture, 2023, 14(1): 60-72.

[16] Fu L, Ding X, Ding Y . Ensemble empirical mode decomposition-based preprocessing method with Multi-LSTM for time series forecasting:a case study for hog prices[J]. Connection Science, 2022, 34(1): 2177-2200.

[17] Ye K, Piao Y, Zhao K, et al. A heterogeneous graph enhanced LSTM network for hog price prediction using online discussion[J]. Agriculture, 2021, 11(4): 359- 359.

[18] Sun F, Meng X, Zhang Y, et al. Agricultural product price forecasting methods: A review[J]. Agriculture, 2023, 13(9): 1671.

[19] Zhou H, Zhang S, Peng J, et al. Informer: beyond efficient transformer for long sequence time-series forecasting[C]. AAAI, 2020, 35(12): 11106-11115.

[20] 王杰, 董国奥, 李俊清. 基于STL-Informer-ARIMA组合模型的猪肉价格预测方法研究[J]. 山东农业大学学报(自然科学版), 2024, 55(03): 367-375.

[21] Zeng A, Chen M, Zhang L, et al. Are Transformers effective for time series forecasting? [C]. AAAI, 2023. 37(9): 11121-11128.

[22] Li Z , Qi S , Li Y , et al. Revisiting long-term time series forecasting: An investigation on linear mapping[J]. arXiv:2305.10721, 2023.

[23] Liu Y, Hu T, Zhang H, et al. iTransformer: Inverted transformers are effective for time series forecasting[C]. ICLR, 2024,pp 1-25.

基金资助

山东省自然科学基金项目(ZR2024MF120)

泰安市科技创新发展项目（政策引导类）(2023NS135)

泰安市科技创新发展项目（政策引导类）(2023NS106)

泰安市科技创新发展项目（政策引导类）(2023ZC485)