A preliminary investigation of overfitting in evolutionary driven model induction: implications for financial modelling

摘要: This paper investigates the effects of early stopping as a method to counteract overfitting in evolutionary data modelling using Genetic Programming. Early has been proposed avoid model overtraining, which shown lead significant degradation out-of-sample performance. If we assume some sort performance metric maximisation, most widely used training criterion is moment within learning process that an unbiased estimate begins decrease after strictly monotonic increase through earlier iterations. We are conducting initial investigation on Programming symbolic regression and financial modelling. Empirical results suggest above increases extrapolation abilities models, but by no means optimal training-stopping case real-world dataset.