Pattern-based causal relationships discovery from event sequences for modeling behavioral user profile in ubiquitous environments

This paper presents a novel and practical model for behavioral user profile modeling using causal relationships. In this model, causal relationships, which represent the influence among variables, are discovered from event sequences representing users behaviors, and used for modeling behavioral user profiles. Our model first discovers significant patterns using probabilistic suffix trees, and then discovers pattern correlations using a new sequence clustering algorithm and a modified version of the normalized mutual information (NMI) measure. Causal relationships between the significant patterns are then discovered using the transfer entropy approach. These relationships are used to construct the causal graphs of activities to generate user profiles. Through extensive experiments over a variety of datasets, we empirically demonstrate that these causality-based profiles lead to significant improvement of performance in activity prediction and user identification. We also show that our proposed model is generic and effective in constructing individual user profiles and common profiles for groups of users, in indoor and outdoor environments.