Development of an Integrated Framework using Machine Learning and TOPSIS for Sustainable Supplier Selection in Circular Supply Chain Management

Sustainable supplier selection is an imperative part in the process where traditional supply chains are transforming into circular supply chains (CSCs) with emphasis on sustainability, resource efficiency, and minimal waste. The supplier selection methods which are traditional are often based on expert opinions and static evaluation frameworks which might be inappropriate to handle dynamic and intricate supply chain environments such as CSCs. In this research, this challenge is addressed by integrating Machine Learning (ML) with the TOPSIS to select sustainable suppliers in the circular supply chain management (CSCM). The ML technique, Random Forest, is utilized in the proposed method to perform feature importance analysis. The various selection criteria are given weights based on the feature importance to minimize bias and enhance accuracy. Suppliers are then evaluated and ranked based on the weights given. The evaluation is based on adherence to the circular economy principles and criteria such as cost, delivery performance and quality which are traditional. To validate the proposed integrated framework, a case study on a food and beverage manufacturing industry was conducted. Data was collected, including primary data obtained through an online structured questionnaire and secondary data from the company’s historical supplier records. The primary data questionnaires were completed by 22 industry experts of the case company. The secondary data was followed by preprocessing, where data was merged and cleaned, the data was then input into the ML Model to derive values of feature importance. These values were next applied in the ML-TOPSIS model to rank suppliers. An analysis between the ML-integrated TOPSIS and traditional TOPSIS methods was conducted to identify the effectiveness of the proposed framework. Findings indicate the objective, adaptive, data-driven decision-making nature of the ML-integrated TOPSIS model. This model dynamically adjusts supplier rankings based on historical data, improving accuracy and efficiency, unlike in traditional TOPSIS, which is heavily based on static, expert-defined weights and longer processing times. This study addresses the literature gap between the ML techniques and Multi-criteria Decision Making (MCDM) methods in selecting sustainable suppliers for circular supply chains. The study provides a systematic tool for industry participants to improve the supplier selection process and promote global sustainable supply chain management. However, the study is limited by the size and scope of the dataset, which may affect generalizability across industries and regions. Future research can address these limitations by expanding datasets, incorporating diverse industry contexts, and exploring advanced ML techniques to improve decision accuracy and applicability.