Estimation of Distribution Algorithm with Machine Learning for Permutation Flowshop Scheduling with Total Tardiness Criterion

Abstract

This paper studies a permutation flowshop scheduling problem (PFSP) with the objective of total tardiness minimization. An improved estimation of distribution algorithm with machine learning, named ML-EDA, is proposed. This algorithm divides the job permutation into several segments and introduces an external archive to keep elite solutions. A two-layer probability model is then constructed in the ML-EDA, and the statistical learning method is employed to produce the probability that the each of job falls at each segment and the probability that the job falls at each location in the segment. Computational results based on benchmark illustrated that the ML-EDA can obtain better solution than the standard EDA for the permutation flowshop scheduling problem with the objective of minimizing total tardiness.