陕西省是我国生鲜农产品冷链物流布局规划的关键之地。生鲜农产品物流规模庞大，但仍存在很多问题，也缺乏直观有效的数据做支撑。为保证更高的物流效率与经济收益，避免因物流供给不足或者过剩导致资源浪费等现象，本文结合理论与实证研究，对陕西省生鲜农产品物流需求进行预测，并对陕西省物流发展现状提出建议与规划。
(1) 针对生鲜农产品物流需求预测指标体系不完善的情况，结合文献与陕西省物流业发展现状，从经济发展水平、产业结构水平、人文因素水平、物流发展规模水平四个方面对物流需求影响因素进行分析，将其扩展成十三个二级指标。考虑到城镇化进程加快，将城镇居民生鲜农产品消费总量作为生鲜农产品物流需求预测的衡量指标。为探究影响因素与生鲜农产品消费总量的关联度，采用灰色关联分析，选择了六个关联度系数在0.8以上的影响因素作为解释变量进行研究。
(2) 针对陕西省生鲜农产品物流需求统计预测模型不成熟的研究现状，利用陕西省统计年鉴2005-2018年的数据，从多元线性回归角度、灰色预测角度、非线性智能预测角度，分别选取主成分回归PCR模型、多变量灰色MGM(1,N)模型、径向基函数RBF神经网络模型进行预测。从非线性智能预测角度，提出了RBF神经网络的无偏灰度理论预测模型，通过引入参数对灰色GM(1,1)模型进行无偏修正，为进一步增强数据的光滑度和减少随机性，采用幂函数对原始数据转换处理，后将GM(1,1)模型与无偏GM(1,1)模型预测精度进行对比，结果显示无偏GM(1,1)模型对序列数据的预测精度更优，同时结合径向基函数神经网络的高度逼近能力，构建了改进的无偏灰色RBF神经网络模型，其预测误差MAPE控制在了5%以下，优化RBF神经网络输入数据使得预测结果更加准确，极大地提高了其在陕西省生鲜农产品物流需求预测方面的适用性。最后，为降低单项模型的预测误差，引入Shapley值进行模型组合，通过权值的确定极大程度地提高了误差分配的合理性，达到了最优组合的效果。预测结果表明，组合模型能以较高的精度对陕西省生鲜农产品物流需求进行预测，预测值最接近真实值，在单项模型预测基础上基于Shapley值组合模型的预测精度得到了进一步的提升。
(3) 针对陕西省生鲜农产品冷链物流业发展存在的问题，根据预测结果，为陕西省冷链物流业提供理论依据与建议对策，以期相关部门能依据预测结果对生鲜农产品冷链物流进行合理的建设，使之适应陕西省生鲜农产品的多元化需求。
 

~Shaanxi province is an important area for cold chain logistics layout planning of fresh agricultural products in China. Despite massive scale of fresh agricultural product logistics in Shaanxi province, there are still many problems and also lacking of intuitive and effective data to support. In order to ensure higher logistics efficiency and economic benefits, and avoid the waste of resources due to insufficient or excessive logistics supply, this thesis combined with theoretical and empirical research to predict the logistics demand of fresh agricultural products in Shaanxi Province, in the meanwhile, this thesis put forward suggestions and made plan for the development of logistics in Shaanxi Province.
(1) Aiming at the imperfect prediction index system of fresh agricultural products logistics demand, combined with literature and the situation of Shaanxi province logistics industry development, logistics demand factors would be analyzed from four aspects, which were the level of economic development, industrial structure, human factors and logistics development scale level. The four aspects would be expanded into 13 secondary indicators. Considering the acceleration of urbanization, the total consumption of fresh agricultural products of urban residents was taken as the measurement index of fresh agricultural products logistics demand prediction. In order to explore the correlation between the influencing factors and the total consumption of fresh agricultural products, this thesis selected six influencing factors by GRA, which correlation coefficient were above 0.8 as explanatory variables.
(2) In view of the research status quo of immature statistical prediction model of demand for fresh agricultural products logistics in Shaanxi province, this thesis used data in the Statistical Yearbook of Shaanxi Province from 2005 to 2018. From the perspective of Multiple Linear Regression, Grey Forecast, Nonlinear Intelligent Forecast, this thesis selected Principal Component Regression Model, Multi-variable Grey Model and Radial Basis Function neural network model respectively for predicting. For the perspective of Nonlinear Intelligent forecast, this thesis proposed the unbiased gray theoretical forecast model of RBF neural network. The GM(1,1) model was unbiased modified by introducing parameters. In order to further enhance the smoothness of data and reduce randomness, the power function was used to transform the original data. Then, the prediction accuracy of GM(1,1) model was compared with the unbiased GM(1,1) model, and the results showed that the accuracy of unbiased GM(1,1) model was more optimal for sequence data forecast. At the same time, combined the unbiased GM(1,1) model with the high approximation ability of RBF neural network, this thesis proposed an improved unbiased Grey-RBF Neural Network model, and the prediction error MAPE was controlled below 5%. The optimization of RBF Neural Network input data made the prediction results more accurate and greatly improved its applicability in the demand prediction of fresh agricultural products logistics in Shaanxi Province. Finally, in order to reduce the prediction error of single model, Shapley value was introduced for model combination, and the rationality of error distribution was greatly improved through the determination of weight for achieving the effect of optimal combination. The prediction results showed that the combined model could predict the demand of fresh agricultural products logistics in Shaanxi province with high precision, and the predicted value was closest to the real value. Based on the prediction of single model, the prediction accuracy based on Shapley value combined model was further improved.
(3) In view of the problems existing in the development of cold chain logistics industry of fresh agricultural products in Shaanxi Province, according to the prediction results, the thesis provided theoretical basis and suggestions for cold chain logistics industry in Shaanxi Province, relevant departments could carry out reasonable construction of fresh agricultural cold chain logistics according to the prediction results, so that it could adapt to the diversified needs of fresh agricultural products in Shaanxi Province.

[1] 王良杰. 陕西省果品冷链物流需求预测[D]. 山西: 太原理工大学, 2016.

[2] 张继良. 基于组合模型的山东生鲜农产品物流需求预测[D]. 山东: 山东大学, 2021.

[3] 缪小红. 生鲜食品冷链物流研究进展[J]. 物流技术, 2009, 28(2): 26-27.

[4] 林荣辉. 供应链环境下生鲜农产品的冷链物流研究——以山东省为例[D]. 山东: 山东理工大学, 2014.

[5] Abad P L, Aggarwal V. Incorporating transport cost in the lot size and pricing decisions with downward sloping demand[J]. International Journal of Production Economics, 2005, 95(3): 297-305.

[6] Shuang Z, Hu Q, Wang D. Research of fresh agricultural products logistics vehicle optimization[J]. International Journal of Intelligent Information Processing, 2011, 2.

[7] Al Theeb N, Smadi H J, Al-Hawari T H, et al. Optimization of vehicle routing with inventory allocation problems in cold supply chain logistics[J]. Computers & Industrial Engineering, 2020, 142: 163.

[8] Gupta S, Haq A, Ali I, et al. Significance of multi-objective optimization in logistics problem for multi-product supply chain network under the intuitionistic fuzzy environment[J]. Complex & Intelligent Systems, 2021, 7(4): 2119-2139.

[9] Silva C A, Sousa J M C, Runkler T A, et al. Distributed supply chain management using ant colony optimization[J]. European Journal of Operational Research, 2009, 199(2): 349-358.

[10] Yanling W, Deli Y, Guoqing Y. Logistics supply chain management based on multi-constrained combinatorial optimization and extended simulated annealing[C]//2010 International Conference on Logistics Systems and Intelligent Management (ICLSIM). IEEE, 2010, 1: 188-192.

[11] Wu X Y, Fan Z P, Cao B B. An analysis of strategies for adopting blockchain technology in the fresh product supply chain[J]. International Journal of Production Research, 2021: 1-18.

[12] Lejarza F, Pistikopoulos I, Baldea M. A scalable real-time solution strategy for supply chain management of fresh produce: A Mexico-to-United States cross border study[J]. International Journal of Production Economics, 2021, 240: 108212.

[13] Petr P, Jan C. Demand forecasting in production logistics of food industry[J]. Applied Mechanics and Materials, 2015, 803: 63-68.

[14] Yu M, Nagurney A. Competitive food supply chain networks with application to fresh produce[J]. European Journal of Operational Research, 2013, 224(2): 273-282.

[15] 曹晓宁, 王永明, 薛方红, 等. 供应商保鲜努力的生鲜农产品双渠道供应链协调决策研究[J]. 中国管理科学, 2021, 29(03): 109-118.

[16] 但斌, 王磊, 李宇雨. 考虑消费者效用与保鲜的生鲜农产品EOQ模型[J]. 中国管理科学, 2011, 19(1): 100-108.

[17] 何忠伟, 桂琳, 刘芳, 等. 北京生鲜农产品物流配送业的发展趋势与质量安全[J]. 北京社会科学, 2010, 4: 43-47.

[18] 崔彬. 生鲜农产品质量安全问题: 成因与间接规制路径[J]. 农村经济, 2010(8): 17-20.

[19] 何旭. 基于HACCP的农产品冷链物流质量控制[J]. 农产品加工, 2010, 9: 69-71.

[20] 焦亚冰. 基于RFID技术的物流信息跟踪系统构建[J]. 计算机工程与设计, 2013, 34(10): 3690-3694.

[21] 尹海博, 徐熊, 郭杭. 基于GNSS的物流信息跟踪系统设计[J]. 导航定位学报, 2021,9(06): 96-103.

[22] 王道军. 基于北斗卫星的物流追踪系统研究[D]. 北京: 北京邮电大学,2014.

[23] Yang H, Jing H. Forecasting of fresh agricultural products demand based on the ARIMA model[J]. Guangdong Agricultural Sciences, 2013, 5(7): 855-858.

[24] Huber J, Gossmann A, Stuckenschmidt H. Cluster-based hierarchical demand forecasting for perishable goods[J]. Expert Systems with Application, 2017, 76(Jun.): 140-151.

[25] Peng Yan, Lin Zhang, Zhiyun Feng, et al. Research on logistics demand forecast of port based on combined model[J]. Journal of Physics: Conference Series, 2019, 1168(3).

[26] Salais-Fierro T E, Martínez J A S. Demand forecasting for freight transport applying machine learning into the logistic distribution[J]. Mobile Networks and Applications, 2022: 1-10.

[27] Guo H, Guo C, Xu B, et al. MLP neural network-based regional logistics demandprediction[J]. Neural Computing and Applications, 2021, 33(9): 3939-3952.

[28] Ma H, Luo X. Logistics demand forecasting model based on improved neural network algorithm[J]. Journal of Intelligent & Fuzzy Systems, 2021, 40(4): 6385-6395.

[29] He F, Chang J. Combined forecasting of regional logistics demand optimized by genetic algorithm[J]. Grey Systems, 2014, 4(2): 221-231.

[30] Huang L, Xie G, Zhao W, et al. Regional logistics demand forecasting: a BP neural network approach[J]. Complex & Intelligent Systems, 2021: 1-16.

[31] Cheng Z, Juncheng T. Adaptive combination forecasting model for China’s logistics freight volume based on an improved PSO-BP neural network[J]. Kybernetes: The International Journal of Systems & Cybernetics, 2015, 44(4): 646-666.

[32] Liu Z Q, Chun-Gui L I, Chen B. Regional logistics demand forecast based on factor analysis and neural network[J]. Computer Simulation, 2012, 29(06): 359-362.

[33] 何柳. 基于灰色预测GM（1.1）模型的辉山农产品物流产业园需求预测研究[D]. 沈阳: 东北大学, 2007.

[34] 何满辉, 逯林, 刘栓宏. 基于模糊粗糙集与支持向量机的区域物流量预测[J]. 交通运输系统工程与信息, 2012, 12(3): 129-134.

[35] 黄福华, 蒋雪林. 生鲜农产品物流效率影响因素与提升模式研究[J]. 北京工商大学学报(社会科学版), 2017, 32(02): 40-49.

[36] 王新利, 赵琨. 基于神经网络的农产品物流需求预测研究[J]. 农业技术经济, 2010(02): 62-68.

[37] 张言彩, 徐宏峰, 郑艳民. 江苏省“十二五”城镇居民冷链物流需求量预测——基于GM(1,1)灰色模型的测算[J]. 安徽农业科学, 2011, 3936: 22699-22701.

[38] 李义华, 王冲, 文哲, 等. 基于滑动无偏灰色模型的湖南省农产品冷链物流需求预测[J]. 中南林业科技大学学报, 2021, 41(08): 161-168.

[39] 王道平, 李锋, 程蕾. 我国农产品物流模式的实证研究——基于各省市的聚类分析法[J]. 财经问题研究, 2011(02): 108-113.

[40] 李夏培. 基于灰色线性组合模型的农产品物流需求预测[J]. 北京交通大学学报(社会科学版), 2017, 16(01): 120-126.

[41] 黄建华, 陈严铛, 卢箫扬. 基于ARIMA-PCR模型的福建省物流需求预测[J]. 武汉理工大学学报(信息与管理工程版), 2019, 41(06): 579-585.

[42] 王晓平, 闫飞. 京津冀农产品冷链物流需求影响因素及预测模型研究[J]. 福建农业学报, 2018, 33(08): 870-878.

[43] 冯永岩. 北京市农产品冷链物流需求预测研究[D]. 北京: 华北电力大学(北京), 2016.

[44] 孙剑青. 北京市物流需求预测研究[D]. 北京: 北京交通大学, 2016.

[45] 中国报告大厅. 冷链物流行业现状分析[EB/OL]. http://www.chinabgao.com/, 2021-01-27.

[46] 张卫. 美国农产品冷链物流现状[J]. 中国食品, 2015 (16): 37-39.

[47] 中国产业研究院. 冷链物流市场规模预测2020冷链物流行业供需现状及市场前景趋势分析[EB/OL]. https://cidi.sufe.edu.cn, 2020-06-17.

[48] 张琳. 价值链视角下生鲜农产品冷链流通模式研究[J]. 改革与战略, 2017, 33(7): 4.

[49] 毕娅. 湖北省城乡食品冷链物流系统需求预测: 理论, 实践与创新[M]. 武汉: 武汉大学出版社, 2016.

[50] Joseph Sussman. Introduction to transportation systems[M]. US: Norwood Artech

house, 2000: 25.

[51] James R.Stock, Douglas M.Lambert. Strategic logistics management[M]. Boston, MA: McGraw-Hill/Irwin, 2001: 5.

[52] 霍蕾. 水产品冷链物流需求预测研究[D]. 北京: 北京交通大学, 2014.

[53] 周名煜, 谢宁, 王承民. 基于灵敏度和灰色关联度的配电网运行方式变权重评估方法[J]. 电力系统保护与控制, 2017, 45(13): 130-137.

[54] 李敏杰, 王健. 基于RBF神经网络的水产品冷链物流需求预测研究[J]. 中国农业资源与区划, 2020, 41(06): 100-109.

[55] 李瑞, 张悟移. 基于RBF神经网络的物流业能源需求预测[J]. 资源科学, 2016, 38(03): 450-460.

[56] 张晓瑞, 方创琳, 王振波, 马海涛. 基于RBF神经网络的城市建成区面积预测研究——兼与BP神经网络和线性回归对比分析[J]. 长江流域资源与环境, 2013, 22(06): 691-697.

[57] 麻秀范, 余思雨, 朱思嘉,等. 基于多因素改进Shapley的虚拟电厂利润分配[J]. 电工技术学报, 2020, 35(S2): 585-595.

[58] 刘双花. 基于改进的Shapley值法对我国清洁能源消费的组合预测[J]. 数学的实践与认识, 2021, 51(08): 203-213.

[59] 王斌会. 多元统计分析及R语言建模[M]. 广州: 暨南大学出版社, 2010: 157-162.

[60] 陈鹏宇, 段新胜. 基于离散指数函数优化GM(1,1)模型的重新优化[J]. 三峡大学学报(自然科学版), 2010, 32(01): 88-91.

[61] 李树, 王丰效. 基于蚁群算法的多变量MGM(1,N)组合预测模型[J]. 数学的实践与认识, 2021, 51(14): 41-47.

[62] 吴霞. 基于粗糙集——神经网络的入侵检测系统的研究[D]. 武汉: 武汉理工大学, 2009.

[63] 刘备, 任栋. 基于小波变换与RBF神经网络的GNSS水汽值预测研究[J]. 大地测量与地球动力学, 2021, 41(12): 1216-1218.

[64] 吉培荣, 黄巍松, 胡翔勇. 无偏灰色预测模型[J]. 系统工程与电子技术, 2000(06): 6-7+80.

[65] 李群. 灰色预测模型的进一步拓广[J]. 系统工程理论与实践, 1993(01): 64-66.

[66] 杨杰, 翁文国. 基于改进无偏灰色模型的燃气供气量的预测[J]. 清华大学学报(自然科学版), 2014, 54(02): 145-148.

[67] 丁松. 灰色预测模型优化及其应用研究[D]. 南京: 南京航空航天大学, 2018.

[68] 张玮玮. 基于聚类分析的BP神经网络短时交通流预测方法研究[D]. 重庆：重庆邮电大学, 2016.

[69] 黄凯, 王健. 我国生鲜农产品冷链物流需求预测分析: 基于最优组合模型[J]. 武汉理工大学学报(信息与管理工程版), 2020, 42(06): 524-529.