Third-party reverse logistics provider selection: A computational semantic analysis-based multi-perspective multi-attribute decision-making approach

Abstract

In the era of circular economies, governments and consumers are increasingly aware of environmental protection, which encourages enterprises to devote more attention to reverse logistics (RL). However, the limited resources and technical limitations of most manufacturing companies have motivated them to outsource their RL activities to professional third-party RL providers (3PRLPs). Optimal 3PRLP selection is instrumentally valuable in RL outsourcing practices because it has the potential to increase enterprises’ economic profitability and to improve their long-term development. Generally, 3PRLP selection is treated as a multiple-attribute decision-making (MADM) problem. To this end, this paper aims to build a multi-perspective MADM (MPMADM) framework to offer systematic decision support for enterprises to select the optimal 3PRLPs. Attribute assessments in the proposed framework take the form of generalized comparative linguistic expressions (GCLEs), which can be transformed into hesitant fuzzy linguistic term set (HFLTS) possibility distributions with semantic analysis in order to enhance information quality and reliability. Expert weights are then assigned in the use of an optimization model based on the correlation consensus measurement. Afterwards, the two-stage aggregation paradigm for computing with HFLTS possibility distributions is used to gather assessments at expert and attribute levels to compile overall assessments of each alternative 3PRLP. Compared with existing studies, our proposal considers environmental and social sustainability for attribute system establishment and introduces GCLEs for 3PRLP selection, which offer greater flexibility for experts to articulate their evaluations. In addition, the two-stage aggregation paradigm eliminates distortion and loss of information and provides decision makers with the capability to control the outcome’s precision. Moreover, the proposed expert weight determination approach is conducive to generating reliable weight vectors. Several illustrative examples, sensitivity analysis, and comparative analysis further demonstrate the flexibility and practicability of our proposal.