Inferring Cause from Effect - A Weight Prediction Case Study
Recent years have substantiated the efficacy of blended care approaches in various medical fields. Particularly, people with obesity and diabetes can benefit from automated dietary interventions due to the availability of in-app bodyweight tracking and AI-supported meal monitoring.
We propose a causal structural model of weight change dynamics inspired by physiological science. This fully interpretable model allows drawing causal links of meal composition and activities tracked by our users, to weight change. We show how the trained parameter space becomes a utile tool in generating predictions and designing interventions, helping patients reach their weight goals.
Philipp Kanehl is a Bayesian enthusiast and a Staff Data Scientist of the Oviva AG. Before migrating to health-tech he worked as a data science consultant with IBM. Philipp holds a PhD in Theoretical Physics from the TU Berlin.