COMPREHENSIVE EVALUATION OF MACHINE LEARNING REGRESSION TECHNIQUES FOR SHEAR VELOCITY LOG PREDICTION (VS)

Abstract

In addition to their numerous other applications, machine learning (ML) techniques can be employed to address regression problems. Linear and polynomial regression are well-established methods utilized widely due to the simplicity of their numerical implementation. But, when confronted with instances of complex problems, these approaches prove to be inefficient. This article examines the effectiveness of Support Vector Regression (SVR) and Bayesian regression in tackling the difficulties associated with regression in the reconstruction of well logs, with a specific focus on shear-wave logs. Regression analysis was employed to construct models that establish the relationship between target variable (features) and independent variables (features) to estimate the S-wave velocity. The findings of this study demonstrate the efficacy of various machine learning (ML) approaches when applied to regression analysis. The graphical analysis and statistical metrics utilized in this evaluation were implemented to quantify and assess qualitative aspects of the performance, respectively. The goal of this study is to establish the machine learning method that yields the most accurate estimation of the s-wave elastic physical property using the Norne field data set as a case study.