In most business domains and research fields, investigating data sources using machine learning has become a standard procedure. With the increasing demand to explain this investigation type better, the realization matured that running machine learning algorithms constitutes only a small part of the required procedure to achieve robust results. Most importantly, domain experts must be better integrated into the loop of the investigation steps. Although many insights have matured in the presented context, their practical implementations are lagging behind. In the field of mobile health, which constitutes an important part of the broader field of digital health, machine learning related results are still rare. Tangible pipelines that can be used off-the-shelf have been also less presented. However, the application of machine learning approaches can help to analyse mHealth-based data sources effectively. Therefore, a deep insight into how this type of data is actually created is needed. This includes a deep understanding of how mobile application engineering is carried out in general as well as for the medical field in particular. Based on a multitude of developed mobile health apps (30), the organizers of this tutorial have collected more than 500.000 data sets from patients all over the world and in the wild, which provides the data basis for the first part of this tutorial. In this, the participants will be provided with relevant aspects and issues of the mobile application engineering side (e.g., offline vs online features), the created data sources (e.g., data collection procedure), and relevant related technical (e.g., proper APIs) as well as medical aspects (e.g., regulatory aspects). In the second part of the tutorial, the participants will get practical insights into conducted machine learning analyses of the developed apps. For the participants, we will provide a detailed discussion with tangible results on the opportunities as well as the shortcomings when using machine learning on mHealth data sources.

Main topics

• In-situ data collection, mobile health (mHealth), mHealth mobile application engineering aspects, regulatory and interdisciplinary aspects, machine learning

Intended audience

• People with an interest on machine learning analyses on mHealth data

Pre-requisites

Of note, an introduction in relevant parts of mobile health will be provided, while an introduction in relevant machine learning concepts (e.g., outcome metrics) or basic data science aspects (e.g., CRISP-DM, statistical tests) will not be provided. Therefore, a basic foundation in machine learning and statistics would be beneficial, but are not exclusion criteria.

Outcomes

• Provision of basic understanding of relevant concepts in mobile health; i.e., medically as well as technically
• Provision of fundamental pros and cons when using machine learning on medical data that was gathered with mobile technology
• Practical insights into existing solutions in relation to the two aforementioned aspects