AI-Enabled Framework for Fresh Food Supply to Reduce Waste and Enhance Resilience

<p dir="ltr">Food is essential for human sustenance, and the food supply chain has a crucial role in society's well-being and stability. Food safety is one of the most important aspects of 17 United Nations Sustainable Development Goals (UN-SDG) The sustainability of food supply chain systems is gaining increasing attention, particularly after the COVID-19 pandemic. A food supply system that simultaneously prioritizes resilience and minimizes waste is crucial for achieving sustainability and streamlining supply chain operations.</p><p dir="ltr">Despite the significant number of studies proposed to reduce food waste, the Food and Agriculture Organization has reported that 14% of global food is wasted annually across various stages of the supply chain. The Food and Agriculture Organization has further revealed that offering sufficient food to the global population by 2050 presents a significant challenge. Furthermore, the global disruption caused by the COVID-19 pandemic has revealed the limitations of the resilient approaches currently utilized within existing food systems, ultimately highlighting the urgent need to develop resilient and adaptive systems. It has been found that many studies have explored reducing food waste and increasing supply chain resilience as separate objectives; however, there is limited research that has investigated both issues simultaneously.</p><p dir="ltr">Literature studies have emphasized the significant roles that emerging technologies play in reducing food waste, enhancing system resilience, and developing sustainable systems. However, prior research still lacks the practical adoption of emerging technologies to reduce food waste and enhance resilience jointly in food supply chain systems. This thesis develops and validates a novel technology-based framework for future sustainable food supply chains that addresses the issues related to product waste reduction and resilience enhancement for the first time. Particularly, this thesis develops an integrated framework of 4 stages and validates it using bananas as a case study, leveraging the predictive capabilities of machine learning within the demand forecasting field.</p><p dir="ltr">Developing effective demand forecasting is crucial for better planning and ensuring sustainability within food systems. The food industry has received the least attention for building demand forecasting approaches. While some models have achieved accurate predictions, they are not assessed for their impact on waste reduction or resilience enhancement.</p><p dir="ltr">To bridge this gap, this thesis develops an integrated framework in which the output from each stage serves as input to the next one. The first stage performs intensive data processing and analysis on datasets collected from real-world experiments, extracting critical insights about the freshness and remaining shelf life of bananas. The second stage utilizes the processed dataset from the previous stage to train and test two models: (1) a classification model to classify food products based on their freshness quality into three classes, and (2) a multiple linear regression model to predict the remaining shelf-life period. Different from existing literature studies, the processed real dataset and the derived insights are integrated into demand forecasting in the third stage. The third stage develops an ensemble stacking model combining the random forest, support vector regression, eXtreme gradient boosting, long short-term memory models as base learners, and Ridge regression as a meta-learner. The developed model is trained and tested on a simulated dataset combined with the real processed dataset from the previous stage to predict the daily demand for fresh food items for a retailer. Finally, several disruption scenarios are developed and tested to assess the forecasting model’s impact on reducing waste and enhancing resilience through utilizing the predicted demand to inform inventory replenishment orders. The promising results indicate the effectiveness of implementing the integrated framework in developing resilient, dynamic, and flexible systems that consider changing conditions and reduce food waste.</p><p dir="ltr">Overall, this thesis highlights the potential of the developed integrated approach to building a model that can effectively reduce waste and enhance resilience jointly in FSCs. This approach enables decision-makers (e.g., food retailers) to track the remaining shelf-life and freshness quality of food products in real-time and feed this information directly into a demand forecasting model. This approach provides critical and up-to-date information that decision-makers use to mitigate disruptions, reduce waste, and enhance the planning process.</p>