Objective model selection with parallel genetic algorithms using an eradication strategy

Abstract In supervised learning, feature selection methods identify the most relevant predictors to include in a model. For linear models, inclusion or exclusion of each variable may be represented as vector bits playing role genetic material that defines Genetic algorithms reproduce strategies natural on population models best. We derive distribution importance scores for parallel under null hypothesis none features has predictive power. They, hence, provide an objective threshold does not require visual inspection bubble plot. also introduce eradication strategy, akin forward stepwise selection, where genes useful variables are sequentially forced into models. The method is illustrated real data, and simulation studies run describe its performance.