MVLSC Home · Issue Contents · Forthcoming Papers
A Hybrid Approach to Delivery Optimisation Using Machine Learning and Spherical Fuzzy TOPSIS
Cagatay Ozdemir, Sezi Cevik Onar and Mustafa Con
This study presents an innovative approach to optimize delivery success in the online retail industry. The research’s core is a machine learning (ML) model that predicts order delivery success, which is made explainable using SHAP (Shapley Additive exPlanations) values. The main novelty of the study is the transformation of SHAP values into Spherical Fuzzy Sets (SFS) and their integration with the TOPSIS method, creating a powerful decision-making mechanism. This mechanism optimizes monthly delivery plans through two scenarios: multi-package delivery, which increases efficiency, and immediate delivery, which enhances customer satisfaction. ML and SHAP serve as tools in this process, while SFS-TOPSIS integration is the main goal of decision-making. This hybrid approach strikes an optimal balance between prediction accuracy, model explainability, and decision-making flexibility. As a result, this study presents a novel methodology that enables the development of data-driven, understandable, and feasible delivery strategies in the online retail sector.
Keywords: Delivery Optimisation, SHAP-SFS Conversion, Spherical Fuzzy Set, Spherical Fuzzy TOPSIS, explainable machine learning, multi-criteria decision making