Development of machine learning-based TOC estimation model using well logging and sidewall core data for source rock evaluation in the Jeju Basin, South Sea of Korea

Key Points

• The aim is to develop a machine learning model for estimating total organic carbon (TOC) using well logging and sidewall core data.
• Used physical logging data and TOC analyses from sidewall core samples.
• Performed data pairing of physical logging and TOC.
• Selected input variables based on available data, including density, sonic, and gamma-ray logs.
• Trained the model using 113 samples and tested it on 5 samples.
• Compared performance between Random Forest and Gradient Boosting methods.
• Gradient Boosting achieved a coefficient of determination (R²) of 0.84 and an average absolute error of 0.09 wt.% on test data.
• Random Forest performed significantly worse with R² of 0.11 and an average absolute error of 0.19 wt.%.
• The model underestimated TOC in sections without core samples, particularly those with over 1 wt.% TOC.