Abstract
Modern railway has a high speed, heavy load and intensive development trend. It not only brings the opportunities of railway transport capacity and volume, but also makes the data of various types of large-scale continuous growth. In the stage of the development of the railway by the production enterprise to the service enterprise, it is necessary to vigorously promote the application of cloud computing, and to plan the cloud computing framework. We study a parallel support vector machine model based on multi-level SVM, and realize the parallel algorithm in cloud computing environment. The algorithm divides the large training data set into a number of small training sets by Map, and then a new SVM is combined with these small t raining sets. Finally, the data is trained to be a new SVM by Reduce. At the end of the paper, we use the SVM parallel forecasting method to predict the passenger flow of China Railway, and compare the performance of the distributed with that of non-distributed algorithms. Experimental results show that the proposed algorithm has better effect than single machine algorithm in terms of time consumption and classification accuracy. With the increase of nodes, the time consumption is significantly shortened.
