Keenness for characterizing continuous optimization problems and predicting differential evolution algorithm performance

Abstract Fitness landscape analysis devotes to characterizing different properties of optimization problems, such as evolvability, sharpness, and neutrality. Although several features have been proposed, only a few them can be used in practice predictors algorithm performance. In this study, the keenness ( $$\textrm{KEE}_{s}$$ KEE s ) is proposed characterize sharpness fitness for continuous problems predict performance differential evolution algorithm. Specifically, mirror simple random walk designed construct relevance between front back search points sampling. The value each point replaced by specific integer. values set integers with same circumstance are computed feature scalar using cumulative calculation mechanism. results experimental studies various functions demonstrate superiority terms accuracy, reliability, coverage samples. Moreover, has shown excellent practicability application prediction problems. Thus, new within limited prior knowledge unknown problem.