A Comparison of Extreme Gradient and Gaussian Process Boosting for a Spatial Logistic Regression on Satellite Data
摘要
A popular and successful method of obtaining regression models using decision tree learners is XGBoost. However, the method implicitly assumes conditional independence of the predictions given the data and is not statistically efficient for autocorrelated data, as arises in spatial statistics. GPBoost incorporates a Gaussian process in a mixed effects model, and is demonstrated for our remote sensing model to reduce the generalisation error dramatically.