A novel evolutionary algorithm based on even difference grey model

Abstract

Abstract Evolutionary algorithm is an important subdomain of the meta-heuristics as a kind of the most popular optimization technology. Inspired by the grey prediction theory, this paper proposes a novel evolution algorithm based on the even difference grey model. Unlike population-based meta-heuristics with employing mutation and crossover operators to generate trial populations, the proposed algorithm develops a reproduction operator by using the even difference grey model. The novel reproduction operator regards population series as a time series. It firstly transforms an unorder data selected from the population series to a series data with approximate exponential law by using a grey operator. Based on the series data generated, an exponential model is constructed by using the even difference grey model. Finally, the reproduction operator obtains a trail population according to the prediction results of the exponential model. The effectiveness and superiority of the proposed algorithm are demonstrated on CEC2005 benchmark functions and six real-world engineering design problems. Philosophically speaking, the proposed algorithm towards a brand-new point of view to achieve the optimization process by forecasting the evolutionary direction at the macroscopic level, while the conventional evolutionary algorithms manipulate the chromosomes to realize their optimization at the microscopic level. It could be hoped from the algorithm to open a door for other prediction theories to construct new meta-heuristics.