The Analysis And Prediction Of Railway Accident Based On Data-driven Method

Train is an important transport method for our country,and it plays irreplaceable role in boosting GDP growth.Recommendations for the 13th Five-Year Plan for Economic and Social Development points out that China will build 23,000 kilometer railroad in the period of the 13th Five-Year Plan,and it will cost more than 2,800 billion yuan.With the increase of railroad,people pay more attention to railway safety issues than before.Railway safety issues involve many aspects,and the possibility of a railway safety issue becomes larger when train is running faster.The prediction of railway accident of an important aspect of the research of railway safety issue.The prediction of railway accident is based on the analysis of railway accident,and it is very important to predict the railway accident by analyzing current condition and history data,before the railway accident occurred.The main content of this paper are as follows:(1)Research the prediction of railway accident by the method of data driven.There are few research focuses on the uncertain data in railway accident data.In this paper,an improved principal components analysis model and an improved total support vector machine model are proposed to deal with uncertain data in railway accident data.Take the data come from Federal Railroad Administration as sample data,analyzing the model and the experiment shows that the prediction result has been risen from 67%to 81%.(2)Propose a deep learning method to deal with uncertain data in the text of railway accident data.A convolutional neural network model is trained by text data set,which come from Federal Railroad Administration,to predict the risk of railway accident.The experiment shows that the prediction result has been risen by 3%compared to the origin method.(3)Propose a railway data processing framework.The railway data processing framework uses NoSQL database,and implemented by object oriented language and algorithm.As for performance,the railway data processing framework is 30%faster than other data processing framework in transportation.