A genetic algorithm for supplier selection problem under collaboration opportunities. Issue 1 (2nd January 2022)

• Record Type: Journal Article
• Authors: Ben Jouida, Sihem; Krichen, Saoussen
• Is Part Of: Journal of experimental & theoretical artificial intelligence. Volume 34:Issue 1(2022)
• Page Start: 53
• Page End: 79
• Publication Date: 2022-01-02
• Subjects: Supplier selection -- genetic algorithm -- assignment problem -- quantity discount -- coalition formation; Artificial intelligence -- Periodicals; 006.3
• DOI: 10.1080/0952813X.2020.1836031
• Languages: English
• ISSNs: 0952-813X

Abstract

ABSTRACT: In this paper, we propose a collaborative model for the supplier selection for the purchasing activity in the supply chain. The problem addresses a set of firms that try to look for a cost saving configuration to optimise their ordering plans, given a set of suppliers with quantity discounts options. Possible collaborations between firms, modelled as a coalition formation, can be beneficial in the sense that the gathering of their orders generates a cost minimisation regarding the stand-alone situation. We propose the mathematical formulation of firms' collaborative ordering modelled as a cost-dependent assignment problem. The collaborative scenario is viewed as a two independent steps: the first step is based on game-theoretic approach to model possible coalitions of firms and to generate stable coalition structures according to the core concept. Once coalitions are formed, the second step consists mainly on the genetic algorithm is trigged to assign coalitions to shared suppliers. The assignment problem is solved using a specifically designed hybrid genetic algorithm. Experiments, driven on a large test-bed, highlight the effectiveness of the collaboration in handling the ordering activity within the supply chain and the usefulness of hybrid genetic algorithm in solving such supplier selection problems. We show that in all cases the collaborative scenario is more profitable regarding the stand-alone position. The obtained results show that the hybrid genetic algorithm is able to generate good quality solutions in a reasonable run time regarding Cplex results.