Application of Machine Learning to support production planning of a food industry in the context of waste generation under uncertainty

• Main Authors: Alberto Garre, Mari Carmen Ruiz, Eloy Hontoria
• Format: Article
• Language: English
• Published: Elsevier 2020-01-01
• Series: Operations Research Perspectives
• Subjects: Output uncertainty; Waste management; Empirical study; Production planning; Sustainability
• Online Access: http://www.sciencedirect.com/science/article/pii/S2214716019301988

Food production is a complex process where uncertainty is very relevant (e.g. stochastic yield and demand, variability in raw materials and ingredients…), resulting in differences between planned production and actual output. These discrepancies have an economic cost for the company (e.g. waste disposal), as well as an environmental impact (food waste and increased carbon footprint). This research aims to develop tools based on data analytics to predict the magnitude of these discrepancies, improving enterprise profitability while, at the same time, reducing environmental impact aiding food waste management.A food company that produces liquid products based on fruits and vegetables was analyzed. Data was gathered on 1,795 batches, including the characteristics of the product (recipe, components used…) and the difference between the input and the output weight. Machine Learning (ML) algorithms were used to predict deviations in production, reducing uncertainties related to the amount of waste produced. The ML models had greater predictive capacity than a linear model with stepwise parameter selection. Then, uncertainty is included in the predictions using a normal distribution based on the residuals of the model. Furthermore, we also demonstrate that ML models can be used as a tool to identify possible production anomalies.This research shows innovative ways to deal with uncertainty in production planning using modern methods in the field of operation research. These tools improve classical methods and provide production managers with valuable information to assess the economic benefits of improved machinery or process controls. As a consequence, accurate predictive models can potentially improve the profitability of food companies, also reducing their environmental impact.