Master’s thesis presentation by Oskar Karnblad and Nils Nordeman, MPBME

Welcome to the master’s thesis presentation “Causal relationships between food intake and stomach issue – An algorithmic detection using machine learning” by Oskar Karnblad and Nils Nordeman, MPBME

When? 9:00, Wednesday, 14 June 2017

Where? Landahlsrummet (room 7430), Hörsalsvägen 11, 7th floor

Examiner: Tomas McKelvey

Irritable Bowel Syndrome is the most common functional disease related to the bowel. It is classified as one of the most common diseases in the world with 11.2% of the global population suffering. Therefore, an accurate tool as an aid will have a major societal impact. In this thesis, algorithms for identifying causal relationships between food intake and stomach issues from synthetically generated data and patient’s self-recorded journals were investigated as the primary aim. The thesis was confined to an investigation of algorithms appropriate for small datasets. Algorithms considered appropriate were members from the following families of algorithms: regression analysis, ensemble learning, support vector machines and Bayesian statistics. The results were obtained by running each algorithm on the same datasets and performing averaging. The study found that the beta-binomial hierarchical model acquired the highest average performance for all metrics considered when selecting symptom intolerances from synthetic data. However, due to the unknown symptom generating behavior of users, the limitations of the model may affect the performance significantly. We believe that utilizing the hierarchical model in combination with another algorithm may be useful for the analysis of the available datasets.