Design And Implementation Of The Parallel Computing In The Railway Goods Traveling Time Forecasting Based On Big Data Spark Method

Railway goods transit time is an important index to evaluate whether the railway transportation mode is competitive in the market.It is of great significance to estimating the railway goods transit time scientifically.It will improve the railway freight transportation service quality,increasing railway traffic volume,and improve the competitiveness of railway freight transport in the transportation market.Railway goods traveling time is a very important part of the total time of railway goods transit time.Due to the existence of many factors like the specialty of railway freight transportation,to many transportation operations process,uncertainty of transportation mode and wagon flow path,the current prediction methods used in the actual production of railways and related theoretical research methods are difficult to estimate the railway goods traveling time accurately,and the accuracy and precision of the prediction results were generally low,which directly affects the forecast accuracy of the total time of railway goods transit time.This paper aims to forecast the railway goods traveling time more accurately.Firstly,it analyses the domestic and foreign research status on the prediction of railway goods traveling time.Then,it points out the existing problems and deficiencies in the method for predicting railway goods traveling time.The paper also combines with railway freight transportation operation process,analyzing the difficulties in predicting the railway goods traveling time.Then,with the help of railway transportation information integration platform storing massive historical message data and real-time data,the paper constructs the forecasting Agent of the railway goods traveling time from two aspects of data processing and time prediction by using Agent Technology.In order to predict the real dynamic forecasting of the railway goods traveling time between any OD sites,the paper constructs a universal forecasting model of the railway goods traveling time based on Agent and the historical data collection by the normal distribution theory and also designs a forecasting algorithm of the railway goods traveling time based on Agent.Then,an experimental case designed to verify the availability of the model.Next,aiming at the large amount of data needs to be processed and the high computational difficulty in the prediction model,the paper carries out the parallel computing design and implementation for the predicting model by Spark.A detailed design of the generally module and five basic modules of forecasting railway goods traveling time of parallel computing was also carried out.Finally,the paper sets up a big data platform to serve the prediction of the railway goods traveling time,which installs many big data application components includingHDFS,HIVE,Spark,etc.Using Scala and Java and other programming methods,a railway prototype system was developed,which is designed to predict the railway goods traveling time.Several experiments have been carried out to verify the effectiveness and practicability of the model and algorithm.