Quantitative Models for Forecasting Demand for Perishable Products: A Systematic Review

Demand forecasting impacts business profitability by assisting in decision-making regarding production and inventory levels to meet demand, without the occurrence of product shortages or waste due to excess. The supply chain of perishable products faces a lack of application, and it is common to observe significant losses in these sectors. To understand how demand forecasting for perishable products has been performed in the literature, this study aims to conduct a systematic review to identify the methods, the data required, and how these studies have been evaluated. Although we usually associate perishable products with food, any other product that cannot be stocked, or that becomes obsolete quickly or has a short shelf life can be considered perishable. This study has shown that the most recent approaches involve classical models such as ARIMA, and machine learning models such as Gradient Boosting, long short-term memory (LSTM) and support vector regression (SVR). In addition, these models are characterized by the use of time series supplemented with external data. To evaluate the models, statistical indicators, such as Mean Absolute Percentage Error (MAPE), are used to measure their results and calculate Root Mean Square Error (RMSE). The results presented in this systematic review allow new studies in this area to be developed based on the main approaches in the literature and are also an opportunity to identify new approaches for methods yet to be more systematically explored.