A model and prototype of a proactive multi-type context-aware recommender system

Abstract

With the huge amount of available information and products, the importance of recommender systems has increased. Currently, these systems are integrating the user's context and social information to create context aware recommender systems. In the future, they will use the big amount of information known about the user from the Internet of things (IoT) environment. The IoT environment will provide a lot of implicit and explicit information about the user, thus knowing more about his context, and consequently enables high quality recommendations. In this paper, we will present a model and prototype of a context aware recommender system that recommends different types of items. The recommender systems recommend three different types (Gas Stations, Restaurants, and Attractions) proactively in the IoT environment. A major part of this design is the context aware management system which decides if the context is proper to push a recommendation or not, and what type of recommendations to push. We assume that the context aware management system inputs are derived virtually from the IoT environment and its output is a score that determines if the context is appropriate for a recommendation and identifies the recommendation type. A neural network was used to do the reasoning of the context. The results of 7000 random contexts were tested. For an average of 93.5% of them, our trained neural network generated correct recommendation types in the correct times and contexts. A user's acceptance survey has been conducted on 50 users; it shows high interest and satisfaction in such application.