Logistic Regression for Prospectivity Modeling

The chapter explores the use of logistic regression for prospectivity modeling in geoscience, addressing the challenge of predicting rare events. It delves into the limitations of standard logistic regression and introduces under-sampling and L2-regularization techniques to mitigate these issues. The authors also propose a method for incorporating nonlinearities into the logistic regression model, enhancing its predictive power. The chapter includes computational results and comparisons with neural networks, demonstrating the effectiveness of the proposed approach. Through a detailed algorithm and case studies, the chapter offers practical insights into improving predictive models in geoscience.