Soft behaviour modelling of user communities

A soft modelling approach for describing behaviour in on-line user communities is introduced in this work. Behaviour models of individual users in dynamic virtual environments have been described in the literature in terms of timed transition automata; they have various drawbacks. Soft multi/agent behaviour automata are defined and proposed to describe multiple user behaviours and to recognise larger classes of user group histories, such as group histories which contain unexpected behaviours. The notion of deviation from the user community model allows defining a soft parsing process which assesses and evaluates the dynamic behaviour of a group of users interacting in virtual environments, such as e-learning and e-business platforms. The soft automaton model can describe virtually infinite sequences of actions due to multiple users and subject to temporal constraints. Soft measures assess a form of distance of observed behaviours by evaluating the amount of temporal deviation, additional or omitted actions contained in an observed history as well as actions performed by unexpected users. The proposed model allows the soft recognition of user group histories also when the observed actions only partially meet the given behaviour model constraints. This approach is more realistic for real-time user community support systems, concerning standard boolean model recognition, when more than one user model is potentially available, and the extent of deviation from community behaviour models can be used as a guide to generate the system support by anticipation, projection and other known techniques. Experiments based on logs from an e-learning platform and plan compilation of the soft multi-agent behaviour automaton show the expressiveness of the proposed model.