Thesis subject

Leveraging data driven approaches and machine learning to characterize ultra-processed dietary patterns (MSc)

Despite the accumulating evidence that increased consumption of ultra-processed food has adverse health implications, it remains difficult to decide what constitutes processed food. In this project, we will develop AI models to identify ultra-processed foods in a Dutch database such as Dutch National Food Consumption Survey, 2019 – 2021.

Short description

There is increasing attention on the prevalence of ultra-processed food and beverage consumption and associated negative health outcomes including poor cardiometabolic outcomes. The prevailing measurement approach for ultra-processed foods, the Nova classification, has been key to establishing the existing body of knowledge. Despite the strong body of evidence linking ultra-processed foods with poor health outcomes, translation to policy solutions and public health messaging is still limited. One challenge to effective translation is the reliance on the subjective identification of these products. Given the ubiquity of ultra-processed foods globally and prevalence of consumption in the Netherlands, a more refined measurement approach that recognizes variation between ultra-processed products, would be of public health value. This series of objectives aims to address exactly that.


Objectives

  1. To conduct a comprehensive literature review and data collection on existing studies related to ultra-processed foods and Artificial intelligence.
  2. To develop AI models to identify ultra-processed foods in a Dutch database (Dutch National Food Consumption Survey, 2019 – 2021)
  3. To compare the concordance and the discordance in identifying ultra-processed foods between the different machine learning approaches.
  4. To identify patterns in ultra-processed food intake using different data driven approaches, and to compare between these approaches and the a-priori approach.

    Tasks

    The work in this master thesis entails:

    • Literature review: Conduct a review of existing research studies related to ultra-processed foods and Artificial intelligence. This will provide a foundation of knowledge and identify research gaps.
    • Data collection and preparation: Gather data relevant to identify ultra-processed foods. Organize and preprocess the data for analysis.
    • AI models development: Design and develop an AI models to identify ultra-processed foods in a Dutch database (Dutch National Food Consumption Survey, 2019 – 2021) and compare the concordance and the discordance in identifying ultra-processed foods between the different machine learning approaches.
    • Identify patterns in ultra-processed food intake using different data driven approaches
    • Compare between AI approaches, data-driven approaches and the a-priori approach.
    • Results reporting and documentation: Prepare a comprehensive report summarizing the research methodology, results, and conclusions.


    Literature

    Requirements

    • Courses: Programming in Python (INF-22306), Data Science Concepts (INF-34306) or Machine Learning (FTE-35306)
    • Required skills/knowledge: Nutrition and health, Machine Learning

      Key words: Artificial Intelligence, food and nutrition, nutrition and health

      Contact person(s)

      Dr. Yamine Bouzembrak (yamine.bouzembrak@wur.nl)

      Neha Khandpur (neha.khandpur@wur.nl)

      Prof. Bedir Tekinerdogan (bedir.tekinerdogan@wur.nl)