Prediction of Subway Vibration Values on the Ground Level Using Machine Learning

This research aims to study the possibility of predicting vibration levels from nearby metro lines using machine learning algorithms. As part of the study, a set of measurements was performed at various construction sites in a certain city, and through the analysis of the data obtained; the influential factors of vibration level were selected. The vibration levels predictions were conducted by building of the classification and regression tree, the random forest and the artificial neural network models. The obtained results of the application of these models are presented in this article; a comparative analysis of the predictions of the algorithms was made, on the basis of which it was established that the artificial neural network demonstrated the best performance in predicting the vibration level in the octave of 31.5 Hz ( R 2 = 0.914, MAE = 2.176, MAPE = 0.031). It is indicated that machine learning algorithms have good fitting and generalization ability and can be used at different stages of infrastructure design.