Development of Price Forecasts for Agricultural Stochastic Planning Models

Traditional supply chains (SCs) for fresh fruits and vegetables (FFVs) are not flexible enough to accommodate sudden shifts in demand and, at times, are incapable of reacting rapidly to disruptive events. One of the root causes of this problem is that there is a lack of data-driven planning to aid decision-making across the SC to better match supply and demand. To address this challenge, works that focus on designing tactical planning optimization models for the FFV SC have emerged. These models help to better plan production according to a grower's resources and the needs of the market - as such, they help match supply and demand. All these models rely on the accuracy and quality of the input data - especially market prices, which is one of the key input parameters. Unfortunately, because publicly available price data in the US is limited, price estimations are often based on the grower's experience which can lead to biased or incorrect results and have serious implications on the operation of a grower's business. As such, there is a need to develop a framework that precedes the planning models and focuses on providing them with data-driven information. In this work, we provide the basis for this needed framework and develop various data-driven price prediction techniques to satisfy the requirements of different planning optimization models (i.e., deterministic vs stochastic). Case studies are presented to demonstrate the developed methodologies.