The solution of the task of dynamic interpretation of seismic data using machine learning methods

Abstract

This article examines the problem of dynamically interpreting seismic data using machine learning models, which include Extremely Randomized Trees (Extra Trees), Gradient Boosting (GB), and Adaptive Boosting (AdaBoost) for the given problem. The study analyzes some existing solutions of the problem and describes the advantages of these machine learning models. Accuracy is estimated using the root mean square error metric. The authors found that dynamic interpretation and prediction of seismic data using these machine learning methods had not been extensively explored in research on related topics, which became the main focus of the study. The article formalizes the use of the mentioned models and highlights features and advantages for the given problem. Several common machine learning methods were investigated to find functional relationships between input parameters. Computational experiments were conducted to evaluate their applicability and compare the algorithms. The results show that the Extra Trees method is the most suitable for practical use for the given problem, as it demonstrates the highest accuracy in determining functional relationships and dynamic interpretation.