Predicting the Sound Speed of Seafloor Sediments in the East China Sea Based on an XGBoost Algorithm

Based on the acoustic and physical data of typical seafloor sediment samples collected in the East China Sea, this study on the super parameter selection and contribution of the characteristic factors of the machine learning model for predicting the sound speed of seafloor sediments was conducted using the eXtreme gradient boosting (XGBoost) algorithm. An XGBoost model for predicting the sound speed of seafloor sediments was established based on five physical parameters: density (rho), water content (w), void ratio (e), sand content (S), and average grain size (M-z). The results demonstrated that the model had the highest accuracy when n_estimator was 75 and max_depth was 5. The model training goodness of fit (R-2) was as high as 0.92, and the mean absolute error and mean absolute percent error of the model prediction were 7.99 m/s and 0.51%, respectively. The results demonstrated that, in the study area, the XGBoost prediction method for the sound speed of seafloor sediments was superior to the traditional single- and two-parameter regressional equation prediction methods, with higher prediction accuracy, thus providing a new approach to predict the sound speed of seafloor sediments.

Automated summary

This study utilizes the XGBoost algorithm to establish a machine learning model for predicting the sound speed of seafloor sediments based on the data from typical samples collected in the East China Sea. The model incorporates five physical parameters and optimizes the hyperparameters to enhance prediction accuracy. The results demonstrate that the model achieves high accuracy in predicting seafloor sediment sound speed.