iHelp: Knowledge Management and Utilisation in the iHelp Platform

This document summarizes the actions performed under T4.1 “Personalized Health Modelling and Predictions” in the context of WP4 “Knowledge Management and Modelling in the iHelp Platform”. More specifically, D4.1 “Personalised health modelling and predictions I” provides an extended description of the mechanisms and Artificial Intelligence (AI) models that will be implemented for the realisation of personalised health and risk prediction models. The implementation of the ΑΙ algorithms that will be realized during the project’s lifecycle highly depends on the provided datasets on which they are trained. At this phase of the project, the description of the datasets that are going to be used, drive the actions under this Task towards a concrete description and specification of the AI algorithms and models that will be utilized. In addition, this document -the first one out of three versions- encapsulates the necessary and relevant information that was researched from recent bibliography to facilitate the manipulation of the available datasets, once they become available, setting the basis for the design and implementation of the models. More specifically, after an initial analysis of the datasets description, this document focuses on the advances of Machine Learning (ML) in the domain of healthcare, with special focus on the importance of Federated Learning (FL) for models trained from user interaction with tracking devices, as also is realised under the scopes of the iHelp project. Moreover, an initial analysis of pre-processing of data is being described, with focus on initial feature selection and imputation. Finally, in this deliverable are analysed the main concepts behind the models, also in compliance with an analysis of the importance of known and unknown risk factors based on the description of the clinical/primary data.