An integrated machin elearning and QFD method to assess risk mitigation strategies

Introduction and Purpose: Plastic waste management has become one of the most critical environmental challenges of the modern era, requiring efficient and resilient supply chains. The recycling supply chain of plastics is exposed to multiple uncertainties, such as fluctuations in the quantity and quality of recyclable inputs, operational instabilities, and process vulnerabilities that threaten its sustainability. In this context, identifying and prioritizing risk factors and formulating preventive strategies to mitigate them are essential. The main purpose of this study is to develop an integrated framework that combines machine learning (ML) algorithms with the fuzzy Quality Function Deployment (QFD) technique to identify key risk factors and prioritize preventive strategies for risk reduction in the plastic recycling supply chain.
Methodology: This research adopts a data-driven approach that leverages machine learning algorithms for risk assessment. The case study was conducted at Northe Shimi Plast Company, one of the largest plastic recycling complexes. The study population consisted of seven industrial experts, each with more than three years of practical experience in plastic waste recycling operations. Through a systematic literature review and expert validation, eleven risk factors and eight preventive strategies were identified. Feature importance techniques from decision-tree and random forest algorithms were employed to calculate the relative weights of risk factors. These weights were then integrated into the fuzzy QFD framework to rank the preventive strategies. Data analysis and model implementation were carried out using MATLAB software.
Findings: The machine learning analysis revealed that input material risks are the most critical threats in the plastic recycling supply chain, followed by recycling process risks and health and safety risks. The fuzzy QFD analysis further indicated that buyer–supplier collaboration represents the most effective preventive strategy for risk mitigation, followed by supply chain transparency and supply chain agility. Buyer–supplier collaboration enhances supply chain resilience through information sharing, joint planning, and contingency strategy development. Meanwhile, digital technologies such as the Internet of Things (IoT), RFID tags, and GPS tracking contribute significantly to improving visibility and real-time risk monitoring across the supply chain.
Conclusion: The results demonstrate that the proposed integrated ML–QFD approach provides a powerful, data-driven tool for risk management and decision-making in recycling supply chains. By automating weight estimation and reducing subjective bias, the model improves the precision and efficiency of the decision-making process. Moreover, the interpretability of tree-based algorithms allows managers to understand the logic behind the model’s outputs and apply its insights in real-world operations.The proposed framework not only strengthens risk management capabilities in the plastic recycling industry but also offers transferability to other recycling sectors, including electronic waste, metal, and rubber recycling. Nevertheless, the model’s effectiveness depends on the availability of sufficient quantitative data. Future research is encouraged to expand the proposed approach by integrating large language models (LLMs) for feature identification, applying operations research techniques such as DEMATEL or network analysis to explore interrelationships among risks, and using objective weighting methods like entropy, CRITIC, or SECA to enhance precision. Integrating this framework with intelligent, data-driven decision tools could further advance predictive risk management and support sustainable supply chain development.