Research On The Application Of Machine Learning Theory In Railway Freight Volume Prediction

The accuracy of railway freight volume prediction is an important basis for the relevant departments of the railway for decision-making, and it provides guidance for rational investment in fixed assets, and makes full use of the railway transport capacity, at the same time, when decision makers carry on the development of transportation planning and programming, accurate and timely forecast results is an important prerequisite for successful planning. China’s current market environment is more complex, railway administration’s net profit is not enough to repay bank loans even interest in recent years, the situation has been going on for a period of time, large changes in the industrial structure and the spatial layout also changed the influence factors of railway freight, the relevance and interpretation of macroeconomic and railway freight volume are falling, for this reason, accurate, scientific and rapid railway freight volume forecasting method and provide guidance for the decision of railway authorities seem to be very necessary, introducing fresh blood and using new theoretical methods of railway freight volume forecasting is becoming more and more important, and we have to be fully aware of the use of new theoretical methods must exist the limitations and immature places, need continuous development and improvement.In this paper, the current situation of railway freight volume is analyzed, the paper confirm in the current environment, the necessity of new railway freight volume forecasting theory and method adapting to the development of the times is put forward. Then this paper summarized prediction methods for railway freight volume, including the traditional forecasting methods and the new methods mentioned in this paper, namely machine learning. Then after analyzing the current economic situation and market environment, find six factors including the finished steel production, the output of raw coal, crude oil processing production, thermal power generation, fixed asset investment and industrial added value growth rate, for the modeling of machine learning algorithms, and selected the monthly value of railway freight related data in 2001-2013 as the original data, using support vector machine, artificial neural network, Bagging algorithm and random forest algorithm for empirical analysis.. In this paper, four kinds of machine learning algorithms were used to forecast the railway freight volume, and programming algorithms and results in the R platform. Finally, the four algorithms were compared, analysis of the pros and cons of each algorithm. According to the application of learning theory in the prediction of railway freight volume, obtained satisfactory forecast result, the four algorithms have advantages and disadvantages, the author believes that the introduction of machine in railway freight volume forecasting in learning theory is a good attempt, the new method is feasible in the forecast of railway freight volume.