Thesis subject

Forecasting algorithms for Food Waste (with Wastewatchers) (BSc / MSc)

Did you know that people tend to eat healthier food on Tuesday, more so than on any other day? In this thesis project you will work closely with Wastewatchers to model food waste.

Short description

Did you know that people tend to eat healthier food on Tuesday? Most people tend to eat vegetables or salads, resulting in the fact that snacks and hot meals are wasted more on Tuesdays, more so than on any other day. Or did you know that when it is rainy and there are many traffic jams, the chance of waste on healthy products increases, while the chance on waste on snacks decreases?
In this thesis project, you will work closely with the Wastewatchers team (and their data) to model food waste data and/or uncover predictors. Wastewatchers are an SME who help food service companies with reducing their food waste by means of data analytics. They’ve automatized food waste management and are developing a forecasting algorithm (e.g. neural network) that helps companies to predict their daily supply and demand. They try to understand human behaviours and, in particular, their (weird) eating patterns. By collecting food waste measurements (1.4 million in total), they are able to run tests and train algorithms to predict why we do the things we do regarding food consumption behaviours.


Objectives

This project will involve working with uncovering predictors. To-date, Wastewatchers have approximately 75 to 85% of all the predictors they need for forecasting consumption behaviours. However, they lack in the last 15% to 25%. The overarching aim of this project is to uncover (un)expected predictors that influence human eating patterns. The following specific considerations may be part of the project:

  1. To continue a topic which is called ‘(in)depent food categories’, to determine the (in)dependent food categories in peoples baskets. (For example, if you buy a hamburger, you’ll buy a hamburger. This is the independent category, sauce, salad, cucumber, pickles and mustard are optional. Those are dependent on the burger and buyer).
  2. Do we react to the weather or do we react to weather forecasts? Wastewatchers suspect that people's eating habits are not influenced by the actual weather, but are programmed by weather forecasts. Can you reflect on or investigate this hypothesis by analyzing weather forecast data and sales/food waste data?

    Tasks

    The work in this thesis entails:

    • Investigating consumer behaviour patterns related to food
    • To assess the solutions available to extract data from the scientific literature (e.g. through an Systematic Literature Review) in a scalable and efficient manner
    • To investigate predictors for modelling food waste behaviours
    • Design/develop/model frameworks or algorithms to improve the prediction of food waste


    Literature


    Requirements

    • Courses: (Optional), Data Science Concepts (INF-34306), Machine Learning (FTE-35306), Big Data (INF-33806)
    • Required skills/knowledge: Basic data science, interest in machine learning or AI, interest in consumer behaviours, etc.

      Key words: Artificial Intelligence, Machine Learning, Statistics, Food Waste

      Contact person(s)

      Dr. Will Hurst (will.hurst@wur.nl)