Abstract
Healthcare systems provide suitable services in different domains to help people in fitting themselves into their best pattern of life. This study is focused on the development of a hybrid reasoning engine called KARE (knowledge acquisition and reasoning engine) which is the core reasoning module of ATHENA (activity-awareness for human-engaged wellness applications) platform(1), carried out at UCLab(2) as a project for promoting active lifestyle. This engine recommends food, mental and physical therapy to the ATHENA users that are based on their personal preferences, historical physical, mental and social health information. In KARE, a hybrid approach is used for reasoning which internally combines the predictions of multiple parallel reasoners into a collective decision. Random Forest, Naive Bayes and IB1 algorithms are used in parallel in each of the reasoner to generate personalized recommendations for the specified service. The predictions of all the individual reasoners are combined using majority voting scheme to enhance the predictive accuracy of the individual reasoner. The proposed hybrid reasoning approach is tested on real world dataset of weight management, collected under the ATHENA project. The accuracy of correct recommendations for food, physical and mental therapies is 98.7%.