Research On Regional Logistics Forecasting Based On Bayesian Deep Network Generalized Linear Model

Along with the development of economic globalization and regional economic integration of the beijing-tianjin-hebei region,the regional economic integration of regional logistics demand level has increased,the rapid development of economy is closely linked with the development of regional logistics,perfect logistics system to provide guarantee for economic development,to mature steady economic growth to promote logistics industry,high efficiency,the two promote each other,complement each other.In order to make the logistics industry drive the economic growth to the greatest extent,it is necessary to build a modern logistics service system,and the forecast of logistics demand is an important basis.This paper selects indicators data related to logistics demand in the Beijing-Tianjin-Hebei region from 1993 to 2019 as research samples.Firstly,data preprocessing and simple descriptive statistics are carried out.Then,the variable data after feature selection is put into the prediction model.Finally,by comparing and finding out the models with high prediction accuracy to form a composite model to predict the future logistics demand in the Beijing-Tianjin-Hebei region.The main research contents are as follows:(1)Pre-processing work,such as missing value supplement and data standardization,has been carried out on the collected logistic related index data over the years in the Beijing-Tianjin-Hebei region to prepare for the selection and prediction of logistic demand characteristics;(2)Visualize the development trend and variables of logistics demand in the Beijing-Tianjin-Hebei region through simple descriptive statistics;(3)The main factors affecting logistics demand in Beijing-Tianjin-Hebei region were selected by using the embedded feature selection method,i.e.the maximum information coefficient method;(4)The GM(1,1)grey prediction model,linear support vector machine model and Bayesian deep network generalized linear model are adopted to forecast the logistics demand;(5)According to the evaluation index of the model,the optimal prediction model-Bayesian deep network generalized linear model is selected,and the composite model formed by the combination of exponential smoothing method and Bayesian deep network generalized linear model is used to forecast,and the logistics demand of Beijing-Tianjin-Hebei region in 2020 and 2021 is predicted to be 297,444 million tons and 336,311 million tons respectively.Based on the research on logistics demand prediction at home and abroad,and combining with the current situation of logistics development in Beijing-Tianjin-Hebei region,this paper studies and forecasts the future logistics demand in Beijing-Tianjin-Hebei region,hoping that the forecast results can provide a basis for the relevant local departments to effectively integrate resources,rationally allocate resources and scientifically guide the development of logistics.The innovation of this paper lies in the use of the maximum information coefficient method for feature selection,which reduces the information redundancy among the influencing factors and improves the accuracy of subsequent model prediction.At the same time,through referring to relevant literature,it is found that both the traditional time series prediction model and the machine learning prediction model have shortcomings.Therefore,in order to compare the differences between the models,this paper not only chooses the traditional time series model and the machine learning prediction model,Bayesian deep network generalized linear model,which is a combination of deep network model and statistical model,is also used to predict the logistics demand in Beijing-Tianjin-Hebei region,which greatly improves the accuracy of the logistics demand prediction model.