Today, I gave a presentation at an IEEE EMBS (Engineering in Medicine and Biology Society) event. My talk was about Predicting Adherence to Ecological Momentary Assessments, and I presented our findings about the predictability of users of a smartphone app filling out questionnaires. Based on smartphone sensor data, we utilized machine learning to try to predict fill-out events. Sensor data only showed moderate predictability, and models based on metadata (features related to past fill-out behavior) achieved a much higher performance.
