Experimental Analysis and Machine Learning of Ground Vibrations Caused by an Elevated High-Speed Railway Based on Random Forest and Bayesian Optimization

With the aim of predicting environmental vibrations induced by an elevated high-speed railway, a machine learning method was developed combining random forest algorithm and Bayesian optimization, using dataset from on-site experiments. When it comes to achieving rapid effective prediction vibrations, there is little research on comparisons between verifications different algorithms, none parameter tuning optimization algorithms. In this paper, field experiment firstly carried out measure ground caused trains running bridge, then vibration characteristics are analyzed in view accelerations weighted levels. Subsequently, three algorithms linear regression, support vector machine, experimental database, their performance discussed. Finally, two models for hyperparameter set further compared. The results show that integrated has higher accuracy than regression machine; excellent high efficiency efficient in-depth parameters can be combined with RF effectively predict railway.