random forest (rf) algorithm

The occurrence of settlement induced by tunneling operations in urban environments is an inevitable phenomenon. The risk of damage to nearby infrastructures and surface structures can be greatly reduced by predicting and controlling this event. This paper uses multiple linear regression (MLR) model and random forest (RF) algorithm to predict the maximum surface settlement (Smax) due to shallow tunnel excavation. Nine input parameters, including the distance of the tunnel center from the ground surface (H), height of the underground water level above the tunnel (W.T), tunnel diameter (D), elastic modulus of soil (E), undrained shear strength of soil (Cu), earth pressure coefficient (K0), unit weight of soil (γ), gap parameter (g), and stability number (N) were selected from 24 data sets related to 14 tunneling projects. The MLR and RF techniques were then implemented for predicting Smax. Three performance indicators of coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) were employed for the training and test phases to evaluate the efficiency of the suggested models. The coefficient of determination values for MLR and RF for training data were 0.814 and 0.957, respectively, while for test data were 0.793 and 0.96, indicating that the RF approach is more efficient than MLR. Moreover, the findings reveal that the RF algorithm exhibits lower RMSE and MAE values in both the training and testing phases compared to the MLR method. This suggests that the RF algorithm exhibits reduced error and higher reliability and accuracy when compared to the MLR model. Also, the performance study demonstrates that among the input parameters, the gap parameter (g) and the undrained shear strength of soil (Cu) have the greatest and least influence on Smax, respectively.