Muhammad Uzair Akmal, Master student in Computer Science at the University of Rostock, will defend his Master's thesis on "Recommending Tips that Improve Health: AI-based Exploitation of State Transitions in Health Profiles" on April 26, 2022, at 11:00 am. The thesis was initiated with Vilua Healthcare GmbH.
Supervisors and reviewers were Fabienne Lambusch and Prof. Michael Fellmann from the University of Rostock. Supervisor at the company was Dr. Jana Hapfelmeier.
The defense will take place in a virtual ZOOM room. All interested parties please contact Fabienne Lambusch (fabienne.lambuschuni-rostockde) by email to submit dial-in data.
Abstract:
Preventive healthcare is very important to recognize problematic health situations even before individuals are affected that much. With this, diseases can be prevented or treated early. Today’s technology can educate people about preventive healthcare and assist in changing behavior. For example, mobile devices like smartphones allow us to continuously track health data and get insights into our own health status. Vilua Healthcare GmbH works on providing personalized health recommendation services concerning Occupational Health Promotion. They already developed a system that collects the health behavior information of users and trains a machine learning model to extract a good recommendation. In this, an automaton-based model has been used that shows, which health states of persons follow each other. Moreover, information about the health actions, which lead to a transition from one health state to another, is given. The aim of this thesis is to improve the existing recommendation selection algorithm by integrating direct feedback of users and importance against recommendations to select the ”best” recommendations from the system. However, there are different selection criteria between which to choose for ”best” recommendations. So, a comparison of which methods fit best to which situations/data sets is missing as it is complex to investigate. This master thesis evaluates the advantages and drawbacks of these methods which will be used for finding the ”best” recommendations. First, a literature review on possible approaches for health recommendations will be conducted considering different recommendation systems in the healthcare domain. A study protocol is developed and implemented that compares proposed methods to select a recommendation. This also includes the definition and creation of synthetic data sets to explore where the proposed methods worked well and where is room for improvement. In the end, an adaption of the Hybrid Recommendation method (a combination of the Appealingness-based method and Importance-based method) is adapted which has the highest accuracy for selecting ”best” recommendations.