Multi-user context inference based on neural networks

Pervasive computing and social computing are two major computing paradigms of this decade, which have evolved more or less in isolation from each other. Integrating pervasive systems with social media can enhance the users' experience and enable them to form pervasive communities with others that share similar interests, habits, profile, behaviour, to communicate and interact with them, to socialise and to share their resources in a seamless manner. But to be able to couple the advantages of pervasive computing and social networking, the wealth of user context information available needs to be properly processed, managed and exploited. This imposes several critical requirements regarding reliable context learning and reasoning facilities. This paper presents a multi-user context inference mechanism that is based on Neural Networks and aims to exploit the knowledge available at community level. This mechanism has been evaluated via various experiments and is proven to perform quite well, even in cases of community members, about whom only limited volumes of historical context data are available.