Learning Profiles Based on Hierarchical Hidden Markov Model
This paper presents a method for automatically constructing a sophisticated user/process profile from traces of user/process behavior. User profile is encoded by means of a Hierarchical Hidden Markov Model (HHMM). The HHMM is a well formalized tool suitable to model complex patterns in long temporal or spatial sequences. The method described here is based on a recent algorithm, which is able to synthesize the HHMM structure from a set of logs of the user activity. The algorithm follows a bottom-up strategy, in which elementary facts in the sequences (motives) are progressively grouped, thus building the abstraction hierarchy of a HHMM, layer after layer. The method is firstly evaluated on artificial data. Then a user identification task, from real traces, is considered. A preliminary experimentation with several different users produced encouraging results.
KeywordsSpeech Recognition Viterbi Algorithm Average Error Rate Real Trace Abstraction Hierarchy
Unable to display preview. Download preview PDF.
- 2.Botta, M., Galassi, U., Giordana, A.: Learning complex and sparse events in long sequences. In: Proceedings of the European Conference on Artificial Intelligence, ECAI 2004, Valencia, Spain (August 2004)Google Scholar
- 9.Murphy, K., Paskin, M.: Linear time inference in hierarchical hmms. In: Advances in Neural Information Processing Systems (NIPS 2001), vol. 14 (2001)Google Scholar
- 10.Rabiner, L., Juang, B.: Fundamentals of Speech Recognition. Prentice Hall, Englewood Cliffs (1993)Google Scholar