Skip to Main Content
Input selection is a crucial stage for empirical modeling of complex processes with numerous features. This correspondence proposes a new hybrid method of case-based reasoning and genetic algorithm (GA) to identify significant inputs from a set of features. Case-based reasoning is performed repeatedly on a "leave-one-out" procedure to yield an unbiased error estimate for a hypothesis. This error estimate is then combined with the number of selected attributes to provide an evaluation function for the GA, which serves as a search engine to find the optimal hypothesis for the input selection problem. Simulation examples and their results are presented to demonstrate the effectiveness of the proposed approach.