Food Waste Prediction in Grocery Stores : time series forecasting by deep learning

Abstract

Food waste has becoming an increasingly important problem globally. Source of waste derives from supply chain, food manufacturing, household, retail stores etc. This thesis focuses on the food waste problem in retail industry and aiming to predict the potential food waste in a grocery store by using deep learning approaches. With a real world dataset from a grocery store in Finland, various deep learning models - MLP, CNN, LSTM, GRU have been trained to forecast the upcoming food waste on product level. The outcome of the experiments have been evaluated by means of calculating the RMSE value as well as a business oriented confusion matrix. The study has demonstrated the capability of the selected deep learning models on predicting the future food waste in retail context.