基于多源数据融合与智能预警的2025年中国大蒜全产业链大数据画像OA
Based on Multi-source Data Fusion and Intelligent Early Warning,Big Data Profiling of the Whole Industry Chain of China's Garlic in 2025
目前大蒜相关研究多聚焦于单因素波动成因或区域局部分析,缺乏多源数据支撑的全产业链系统性刻画.本文旨在构建大蒜全产业链大数据画像,揭示产业运行规律与供需传导机制,为产业精准调控与经营决策提供科学依据.通过融合2015-2025年遥感监测、气象观测、实地调研、市场流通及宏观贸易与行业统计等多源异构数据,搭建覆盖生产供给、库存、加工、消费、国际贸易、成本效益及市场价格的大蒜全产业链数据画像框架;采用随机森林监督分类提取种植空间分布,结合统计分析、相关性分析解析各环节运行特征;构建经粒子群算法优化的变分模态分解-时间卷积注意力网络(VMD-TCAN)价格预测模型,通过贝叶斯优化完成超参数寻优与精度验证.结果表明:2025年全国大蒜种植面积85.53万hm2、总产量1 592万t,9月冷库库存量达505万t,均处于近年高位;国内消费总量1 243万t,出口呈现"量增价跌"态势,产业整体遵循"供给扩张—库存累积—价格下行—效益收缩"的传导逻辑,存在出口附加值偏低、精深加工驱动不足、库存调控能力滞后等核心问题.VMD-TCAN模型可精准捕捉价格多尺度波动规律与外部因素非线性影响,拟合效果优异、泛化能力强;预测2026年金乡大蒜价格在3~6元/kg,全年呈先降后稳、低位企稳的特征.研究可为大蒜全产业链监测预警与高质量升级提供数据支撑与方法参考.
Current research on garlic mostly focuses on single-factor fluctuation causes or localized regional analysis,lacking a systematic characterization of the entire industrial chain supported by multi-source data.This paper aims to construct a big data profile of the entire garlic industry chain,reveal the operating patterns of the industry and the supply-demand transmission mechanism,and provide a scientific basis for precise industrial regulation and business decision-making.By integrating multi-source heterogeneous data from 2015 to 2025,including remote sensing monitoring,meteorological observations,field surveys,market circulation,macro trade and industry statistics,this study establishes a data profiling framework for the whole garlic industry chain covering production supply,inventory,processing,consumption,international trade,cost-effectiveness and market prices.Random forest supervised classification is used to extract the spatial distribution of planting,and statistical analysis and correlation analysis are combined to examine the operating characteristics of each link.A price forecasting model based on the variational mode decomposition-temporal convolutional attention network(VMD-TCAN)optimized by particle swarm optimization is constructed,and hyperparameter optimization and accuracy validation are carried out through Bayesian optimization.The results show that in 2025,the national garlic planting area was 855 300 hm2,total output was 15.92 million t,and cold storage inventory in September reached 5.05 million t,all at high levels in recent years.Domestic consumption totaled 12.43 million t,while exports showed a pattern of increasing volume and declining prices.The industry as a whole follows the transmission logic of'supply expansion-inventory accumulation-price decline-benefit contraction',and there are core problems such as low export added value,insufficient driving force from deep processing,and lagging inventory regulation capacity.The VMD-TCAN model can accurately capture multi-scale price fluctuation patterns and the nonlinear effects of external factors,with excellent fitting performance and strong generalization ability.It is predicted that in 2026,Jinxiang garlic prices will mainly range from 3 to 6 yuan/kg,showing a pattern of first declining and then stabilizing at a low level throughout the year.This study can provide data support and methodological reference for monitoring and early warning of the entire garlic industry chain and for its high-quality upgrading.
周浩男;柳平增;朱浩楠;朱珂;张艳;刘彭;许世卫
山东农业大学信息科学与工程学院,山东泰安 271018||农业农村部黄淮海智慧农业技术重点实验室,山东泰安 271018||山东农业大学农业大数据研究中心,山东泰安 271018山东农业大学信息科学与工程学院,山东泰安 271018||农业农村部黄淮海智慧农业技术重点实验室,山东泰安 271018||山东农业大学农业大数据研究中心,山东泰安 271018山东农业大学信息科学与工程学院,山东泰安 271018||农业农村部黄淮海智慧农业技术重点实验室,山东泰安 271018||山东农业大学农业大数据研究中心,山东泰安 271018山东农业大学信息科学与工程学院,山东泰安 271018||农业农村部黄淮海智慧农业技术重点实验室,山东泰安 271018||山东农业大学农业大数据研究中心,山东泰安 271018山东农业大学信息科学与工程学院,山东泰安 271018||农业农村部黄淮海智慧农业技术重点实验室,山东泰安 271018||山东农业大学农业大数据研究中心,山东泰安 271018山东农业大学信息科学与工程学院,山东泰安 271018||农业农村部黄淮海智慧农业技术重点实验室,山东泰安 271018||山东农业大学农业大数据研究中心,山东泰安 271018中国农业科学院农业信息研究所,北京 100081
农业科技
大蒜产业大数据画像供给预测价格波动成本效益分析模型变分模态分解
garlicindustrybig data profilingsupply forecastingprice fluctuationcost-effectiveness analysismodelvariational mode decomposition
《蔬菜》 2026 (8)
92-106,15
2025年山东省农业重大技术协同推广计划(SDNYXTTG-2025-27).
评论