Machine Learning-Based Prediction of Terrestrial Heat Flow in the Songliao Basin

This study investigates the application of Support Vector Machine (SVM) regression analysis for predicting Terrestrial Heat Flow in the Songliao Basin, offering an innovative method distinct from traditional geothermal prediction techniques. Focusing solely on SVM regression within a comprehensive geological and geophysical dataset, our findings reveal a significant predictive capacity, with the highest Terrestrial Heat Flow estimated at 78 mW• m -2 in the basin's center and the lowest at 40 mW• m -2 on its periphery. This pattern indicates a concentric decrease from the center outward. Despite the promising results, the SVM model's accuracy necessitates further enhancement. Future research directions include integrating SVM with other machine learning models, like random forests, to improve predictions. The analysis also highlights crustal thickness as the most influential factor on heat flow, with the annual average temperature having minimal impact. This study not only demonstrates the SVM model's potential in geothermal exploration but also sets the stage for advanced, hybrid machine learning approaches in future terrestrial heat flow research, contributing significantly to the field of geothermal resource management.