Study On The Early Warning Method Of Railway Freight Market Based On SVM And RBF Neural Network

In recent years, the total amount of whole society transport demand of bulk materials continues to decline, such as coal, ore, steel and so on. This is due to the slowing of national macroeconomic growth and increasing efforts of economic restructuring, industrial transformation and upgrading and the energy conservation, which also results in the significant decrease of bulk supply. Since 2013, a marked decline in railway freight illustrates that the severe situation has been an indisputable fact. Based on the analysis of railway freight market situation and influencing factors, an early warning method of railway freight market was proposed with reference of related research results. It would make a comprehensive evaluation and early warning to railway freight market under’new normal’economy. After all, it was aimed to provide a new way and scientific method for the early warning of railway freight market.Firstly, the meaning of early warning of railway freight market was defined. Simultaneously, the basic flow including clearness of warning meaning, alarm analysis, explore source, pre-alarm and regulation was confirmed. And then, early warning methods and basic theoretical models of railway freight were analyzed. Integrated simulation and model warning method were chosen to judge the alarm degree and build early warning models of the rail freight market.Then, the construction of early warning indicator system of railway freight market was researched. On the basis of the existing research by scholars,41 alternative indicator set was proposed. Among them,31 indicators were selected as correlation calculated by gray correlation method, which were classified into’pressure’,’state’ and ’response’ subsystems according to’Pressure-State-Response’(PSR) model. These three subsystems were corresponded to demand, supply and improvement measures. In addition, weight of each indicator was obtained by entropy method. Furthermore, the calculation of early warning index and division interval of alarm were put forward.Next, early warning method based on support vector machine (SVM) classification method and rolling early warning method based on radial basis function neural network (RBF) were raised for railway freight market. Detailed implementations in MATLAB were also given.Finally, a case study on the early warning method of railway freight market was shown, based on the calculation and analysis of nearly 24-month alarm situation. Classification results through SVM model and prediction results of early warning index were tested respectively by MATLAB. Besides, the accuracy of the two models were high. Then, similar alarm situation of the next three months was forecasted by two models. Furthermore, some control measures were proposed. So that the relevant departments and enterprises can adopt regulatory measures and marketing strategy as soon as possible to promote the stable operation of the railway freight market.