Study On The Logistics Demand Forecasting Of Guangdong Province Based On DTRS-SVM Model

Forecasting the demand of regional logistics is the focus of regional economy and logistics management plan, and it is also the premise of planning the regional economy and logistics activities. The results of the regional logistics demand forecasting can affect the logistics infrastructure, transportation network layout and the future development of the local logistics enterprises. Forecasting the demand of regional logistics is the premise of the development of logistics. After the forecasting of the regional logistics demand, we can familiar with the logistics demand of the local economic activities in time. And we can meet the logistics demand of the local economic activities, and then keep the balance between logistics service and demand. It makes the local economic construction and logistics develop coordinately, and maintain a comparatively high efficiency.There are many papers focused on the logistics demand forecasting at home and abroad. Every scholar uses different models to forecast the local logistics demand. Most scholars forecast the volume of freight and freight turnover. The earliest forecasting methods are average growth rate, grey model and random time series model, which are used in the earliest research. Using single modeling method is difficult to have an accurate forecasting result, whose forecasting accuracy is relatively poor. So in recently research, most forecasting methods are combined with more advanced models. In recent years, scholars begin to use machine learning model to derive a relatively more accurate trend by considering the relationship between the dependent variable and the influencing factors and combining the historical data with machine learning. The most commonly used model is neural network model and support vector machine model.This paper is on the basic of the research of logistics demand forecasting conducted by domestic and foreign scholars. Firstly it summaries the existing problems of current research, concludes the concept and the characteristics of regional logistics, and analyzes of the contents and steps of logistics demand forecast. Secondly, it analyzes the main influence factors of logistics demand, and then constructs a suitable forecasting index system of economic aggregate, industrial structure and local residents’ consumption. Thirdly, after a detailed explanation of the decision of rough set theory, it introduces the attribute reduction algorithm to the forecasting model of this paper, in order to eliminate redundant attributes and optimize the model. Then it explains the theory of support vector machine and its model, and uses the genetic algorithm to optimize the kernel function parameter of the SVM model, and then improves the accuracy of the prediction further on the basis of attribute reduction algorithm. Finally, it uses the established model to forecast the logistics demand of Guangdong Province on MATLAB software. Through the comparison of the regression effect before and after attribute reduction and parameter optimization, it proves that this model is more effective and feasible than a single support vector. This paper provides a new idea and method for regional logistics demand forecasting.