基于贝叶斯概率理论的物流园区选址优化研究

物流园区选址是其规划建设中不可或缺的部分，传统的选址方法以定性分析为主，缺乏对选址结果可靠性的评判标准，为此，引入贝叶斯概率方法开展了物流园区选址规划理论研究。借助EM算法和GeNie软件，构建了贝叶斯网络模型，挖掘出各影响因素间存在的内在因果和逻辑关系，量化各影响因素的显著性，构建科学合理的物流园区选址规划评价指标体系；采用K-means聚类方法建立了物流园区选址模型，使用MATLAB软件对建立的模型进行迭代计算，并引入贝叶斯判别方法对聚类结果进行了可靠性分析；基于贝叶斯网络方法优化了灰色模糊风险评估模型，结合了定性分析和定量计算，得出相应的选址风险系数与评估风险概率，完成了物流园区的选址风险等级评估。研究结果表明：基于贝叶斯方法建立的物流园区选址模型能够充分考虑城市规模、经济社会发展、城市物流量及交通区位等多因素的影响，实现了对物流园区选址结果的量化评价，降低了人为主观因素和客观数据的模糊性对物流园区选址方案的影响，有效提高了物流园区选址规划的科学性和可靠性。以陕西省10个地级市为候选地点验证该选址模型的可行性，结果显示：西安建设物流园区的选址风险系数为4.030 1，处于低风险水平，依据总体风险概率确定低风险水平的可靠性为65.50%，证实了在西安建设物流园区（西安港）风险系数较低，科学可行，且西安港风险评估结论具有较高的可靠性。

Location Optimization of Logistics Park Based on Bayesian Probability Theory

The site selection of a logistics park is an indispensable part of its planning and construction. Conventional site-selection methods mainly focus on qualitative analysis and lack evaluation criteria for the reliability of selection results. In this study, the probability method was used to analyze the site selection and planning theory of a logistics park. By using the EM algorithm and GeNie software, a Bayesian network model was constructed to explore the internal causal and logical relationships among the influencing factors, quantify the significance of each influencing factor, and construct a scientific and reasonable logistics-park site-selection planning evaluation index system. The K-means clustering method was adopted to establish the logistics-park-location model, and MATLAB was used to iteratively calculate the established model. The gray fuzzy risk assessment model was also optimized based on the Bayesian network method. Moreover, the combination of qualitative analysis and quantitative calculation was used to obtain the corresponding location risk coefficient and evaluation risk probability, and the site-selection risk-level assessment of the logistics park was completed. The results show that the site-selection model based on the Bayesian method can fully consider the influence of city scale, economic and social development, urban logistics volume, and traffic location in the selection process; realize the quantitative evaluation of logistics-park-location results; reduce the impact of human subjective factors and objective data ambiguity on site selection; and effectively improve the site selection of logistics parks. Ten prefecture-level cities in Shaanxi Province were selected as candidate sites to verify the feasibility of the site-selection model. The site-selection risk coefficient of Xi'an logistics park was 4.030 1, which indicates a low risk level. According to the overall risk probability, the reliability of determining the low risk level was 65.50%. Thus, the results confirmed that the risk coefficient of constructing logistic parks in Xi'an is low, which is scientifically feasible. Moreover, the risk-assessment result of the Xi'an port has relatively high reliability.