An approach to modeling call response behavior on mobile phones based on multi-dimensional contexts

Due to the popularity of context-aware computing and the rapid growth of the smart phone devices, modeling an individual's phone call response behavior may assist them in their daily activities for managing call interruptions. A key step of such modeling is to discovering call response behavioral rules based on multi-dimensional contexts related to individual's behavior. Currently, researchers use classification rule learners for modeling individual's mobile phone behavior. However, the problem is that such learning techniques produce only rules that include maximal number of contexts albeit ordered by relevance. This results in many rules with low-reliability that decrease the accuracy of the modeling approach. In this paper, we propose an approach (Tmodel) to modeling individual's phone call response behavior utilizing mobile phone data. This approach produces not only general rules that capture individual's behavior at a particular level of confidence with a minimal number of contexts, but also produce rules that express specific exceptions to the general rules when more context-dimensions are taken into account. Experimental evaluation shows that our approach outperforms existing approaches to modelling individual's phone call response behavior based on multidimensional contexts.