Impact of Encoding and Neighborhood on Landscape Analysis for the Job Shop Scheduling Problem

Abstract

In order to improve the performance of metaheuristics, many studies focus on the efficiency of neighborhood operators and on the hybridization of various methods. Though, the solution encoding scheme has an important role in the search space's definition, their choice is usually made without any relevance study nor valuable justification. Thus, the present paper aims at conducting analysis on fitness landscapes generated by a basic job shop scheduling problem testing three encoding schemes combined with three operators to optimize the makespan scheduling criterion. Through the study of three different metrics, we show that the encoding schemes associated to neighborhood operators play a major role in the structure of the fitness landscape.