Research On Railway Freight Volume Forecast Based On Machine Learning Theory

Cargo transportation is one of the core businesses of railway;the transportation business in the national goods transport plays an important role nowadays.The national railway freight center was established in 2013,is facing new development opportunities.And Railway freight volume forecasting future demand is very urgent;therefore,accurate prediction in advance will bring many new resources and new opportunities,many railway companies to the vehicle and the main reference for resource allocation.There are many available method of railway freight volume forecasting so far,but,as the change of the domestic market environment,factors of influence on railway freight volume is also becoming more complicated,the previous method also gradually show their deficiencies and limitations.Therefore,we need to introduce some new method to adapt to the present age development for traffic volume forecast.This paper firstly analyzes the background of domestic freight,put forward the necessity and urgency of the new method;and then analyzed and summarized the domestic and foreign related literature analysis of the current situation of the research;through the analysis of the specific domestic market environment,identified 7 factors,namely: railway goods turnover volume,GDP,employment railway transportation,waterway freight volume and freight volume of highway and railway mileage,the output of raw coal and steel production;then using SVR algorithm,BP neural network,Adaboost algorithm,GBDT algorithm(Gradient Boosted Decision Tree)was predicted,obtained with the freight volume forecasting methods,and through comparison conclusion.In general,Adaboost algorithm stability,high accuracy,and it is an ideal way to predict.