Sistema de recomendaciones consciente del contexto como apoyo a programas de promoción de actividad física

Abstract

Recommender systems (RS) are useful tools for filtering and sorting items and information for users. There is a wide diversity of approaches that help creating personalized recommendations. Context-aware recommender systems (CARS) are a kind of RS which provide adaptation capabilities to the user’s environment, e.g., by sensing data through wearable devices or other biomedical sensors. In healthcare and wellbeing, CARS can support health promotion and health education, considering that each individual requires tailored intervention programs. Our research aims at proposing a context-aware mobile recommender system for the promotion of healthy habits. The system is adapted to the user’s needs, his/her health information, interests, time, location and lifestyles. In this paper, the CARS computational architecture and the user and context models of health promotion are presented, which were used to implement and test a prototype recommender system. Context-aware recommender systems (CARS) are a kind of recommender system that adapts to the current circumstances of the user, providing accurate recommendations about different products, services and/or resources. Contextual information can be obtained from online resources, services, stationary or mobile devices, or wearable sensors. CARS address the fact that users interact with the system within a particular “context", and when the context changes, the preferences may also vary. CARS have huge potential for supporting health promotion programs, e.g., by recommending educational multimedia resources. Examples of CARS are Mopet, Fittle and Empower; they are mobile applications that make recommendations for stretching exercises and strengthening outdoor activities, based on user’s information. Diabeticlink is a mobile application, which recommends videos and articles about exercise and healthy diabetic diet, based on user data and his/her interaction with the system. It refers to other data collected through sensors, and interrelates user’s life styles and risk factors. This Thesis proposes a recommender system as well as user and context models for health promotion CARS. Based on the proposed architecture, we implemented the context aware recommender system CoCARE, able to recommend appropriated multimedia resources on physical activity (PA) and healthy diet (HD).