2026-05-25T03:13:13Zhttps://keep.lib.asu.edu/oai/request

oai:keep.lib.asu.edu:node-1981592024-12-23T18:01:48Zoai_pmh:repo_items

198159 https://hdl.handle.net/2286/R.2.N.198159 http://rightsstatements.org/vocab/InC/1.0/ All Rights Reserved 2024 2026-12-01T11:19:46 216 pages Doctoral Dissertation Academic theses Text eng Hernandez Cruz, Xaimarie Villalobos, J. Rene Runger, George Iquebal, Ashif Gonzalez Ramirez, Rosa Arizona State University Partial requirement for: Ph.D., Arizona State University, 2024 Field of study: Industrial Engineering In recent years, US fresh fruit and vegetable (FFV) supply chains (SCs) have been criticized for lacking timely market information, limiting their ability to adapt to disruptions, contributing to food scarcity risks, and hindering small growers' access to beneficial markets. While planning tools are available to reduce these adverse effects, they depend on historical data and fail to account for real-time market conditions, hindering proactive responses to price fluctuations and other disturbances. This dissertation addresses these challenges by developing a novel market intelligence (MI) layered framework that collects, curates, monitors, and forecasts market information to enable data-driven decision-making and planning in the SC. The framework fuses heterogeneous traditional and non-traditional data to detect market disruptions or opportunities via statistical monitoring techniques. It also leverages novel association-mining methods to identify leading indicators of the market price. The generated insights help growers, and other SC stakeholders anticipate disruptions, giving them enough time to take preventive action while economically benefitting from the market opportunity and reducing food waste. A key contribution is the ability to generate timely insights with a low false-positive rate, improving planning decision-making processes in the SCs that enable the retention of growers in the agri-business by connecting them to beneficial markets. Given the data scarcity of public FFV local market price data in the US, this framework also introduces a novel sequential transformer-based spatial-temporal prediction method. This method first forecasts attributes using data for locations with available information, then interpolates predictions for locations lacking data. The developed method allows the incorporation of numerous auxiliary variables into the modeling process and remains efficient even with a limited number of data-present regions, capabilities lacking in alternative methods. Additional alternative synchronous methods are developed and explored to overcome the inefficiencies of separate spatial-temporal modeling, providing a comprehensive evaluation of their performance and computational efficiency. Addressing the data scarcity challenge through these methods expands the framework's applicability to data-absent regions, benefiting a broader population of growers and consumers. Furthermore, this research advances spatial-temporal modeling literature by developing innovative high-performing model architectures to forecast attributes for regions lacking data. Industrial Engineering Artificial Intelligence Agriculture association mining food supply chains market intelligence price monitoring spatial-temporal modeling Transformers Machine Learning for Market Intelligence of Fresh Produce Supply Chains