Prediction Model and Method of Train Body Vibration Based on Bagged Regression Tree

The vibration acceleration of train body is a key parameter reflecting the running state of train. It is necessary to obtain the acceleration accurately. But the traditional method has low precision. In this paper, a vibration acceleration prediction model and method of train body based on bagged regression tree is proposed. On the basis of GJ-5 to collect a large number of parameters of Guangzhou works section in Guangzhou-Shenzhen II line, Pearson correlation coefficient, Spearman correlation coefficient, and Kendall correlation coefficient are used to analyze the correlation between train body vibration and other detection parameters. Then, the bagging regression tree algorithm is used to establish the prediction model of train body vibration. Finally, the training results are compared with the outputs of the model with multiple linear regression model, support vector machine, and back propagation neural network. According to the evaluation index, the prediction accuracy of the bagged regression tree model is highest compared other three models, which is over 94%.