JMIR Publications

ABSTRACT

An automatic electronic coaching (eCoaching) can motivate individuals to lead a healthy lifestyle through early health risk prediction, customized recommendation generation, preference setting (such as, goal setting, response, and interaction), and goal evaluation. Such an eCoach system needs to collect heterogeneous health, wellness, and contextual data, and then convert them into meaningful information for health monitoring, health risk prediction, and the generation of personalized recommendations. However, data from various sources may cause a data compatibility dilemma. The proposed ontology can help in data integration, logical representation of sensory observations and customized suggestions, and discover implied knowledge. This "proof of concept (PoC)" research will help sensors, personal preferences, and recommendation data to be more organized. The research aims to design and develop an OWL-based ontology ("UiA Activity Recommendation Ontology/UiAARO") to annotate activity sensor data, contextual weather data, personal information, personal preferences, and personalized activity recommendations. The ontology was created using Protégé (V. 5.5.0) open-source software. We used the Java-based Jena Framework (V. 3.16) to build a semantic web application, which includes RDF API, OWL API, native tuple storage (TDB), and SPARQL query engine. The HermiT (V. 1.4.3.x) ontology reasoner available in Protégé 5.x has implemented the logical and structural consistency of the proposed ontology. The ontology can be visualized with OWLViz and OntoGraf, and the formal representation has been used to infer the health status of the eCoach participants with a reasoner. We have also developed an ontology verification module that behaves like a rule-based decision making (e.g., health state monitor and prediction), which can evaluate participant’s health state based on the evaluation of SPARQL query results, activity performed and predefined goal. Furthermore, the “UiAARO” has helped to represent the personalized recommendation messages beyond just “String” values, rather more meaningful with object-oriented representation. The scope of the proposed ontology is limited neither to specific sensor data nor only activity recommendations; instead, its scope can be further extended.