Evaluating the use of mobile sensors in improving the user model in mobile recommender systems

Abstract

Information Systems have influenced decision making through recommender systems in several contexts. Such systems have brought about a learning process whereby systems have become able to adapt themselves better to an environment and offer more accurate information to the user. The aim of this paper is to evaluate the possibility of combining various mobile sensors available in consumer electronics in learning systems, thereby providing a more accurate representation of a user’s interests. Such information would successively be used to provide recommendations. To achieve this aim, two mobile applications have been developed. The first application gathered sensor data during simulations of basic user movement in three scenarios. The data gathered successively made up the basis for a second application that consisted of a prototype user profile builder in a tourism context. The results obtained show that it is possible to create a more accurate profile although there is considerable room for improvement. This is so because in some cases, the system registered interest incorrectly for places that were in close proximity to the actual place visited. Similar inaccuracies were attributed to the quality of the sensors and the complexity in devising an algorithm capable of capturing all possible combinations user movement and device handling.