An Internet of Things (IoT) Approach for Automatic Context Detection

Context-aware systems are intelligent systems that recommend users with adaptive service choices which are appropriate to their profile preferences and contextual situations. However, the key challenges in these systems are extracting dynamic context information and updating user's profile information. Recently by the popularity of IoT devices as well as the availability of the wireless network, a massive volume of data is available to gain useful insights into user profile which contains dynamic contextual information. IoT sensors continuously generate enormous amounts of data. However, in order to add value to raw sensor data, we need to understand them and use them in a context-aware computing platform. In this article, an adaptive context-aware service composition system is proposed using an IoT based social network dataset to find users' preferences, improve quality of experience, and boosts the personalization level of services. This paper proposes an 10T-based context-aware service composition to seamlessly find adaptive services offered by a service provider at anytime and anywhere. Firstly, the proposed system considers the dynamic social profile of the user and clusters their preferences while other systems take only static user interests into account. Secondly, this system increases the level of personalization by extracting context information from a user's IoT devices. Finally, the adaptive service composition provides users with various context-aware recommendation including movies, music, venues, radio stations, and events simultaneously. To evaluate the efficiency of the proposed system, we target some of the smart objects such as smart watches/bands, smartphones/tablets, Google Home, and personal computers for detecting several contexts (e.g location, time, user activity, interest and e.t.c.). Experimental evaluations show that the proposed algorithm can improve accuracy and personalization of filtered items with compare to other widely-applied conventional filtering algorithms.