The paper presents an approach to providing advice on health related quality of life to patients with congestive heart failure, using predictive models built from telemonitoring data. First, by combining machine learning algorithms, feature construction, feature selection and expert knowledge, we built a set of predictive models. We then identified which of the features present in the models can be changed by the patients themselves with an appropriate intervention and modelled the association between them and all the other features using linear models. At the end, by using multi-objective optimization, we found the inimum necessary changes of the modificable features that improve the patients' feeling of health. This way we can provide a set of appropriate advices for patients. Thefindings mostly correspond to the current medical knowledge, although some may represent new insights.
Predictive models to improve the wellbeing of heart-failure patients
Saturday, June 24, 2017
Artificial Intelligence in Medicine conference, Workshop on Advanced Predictive Models in Healthcare