Evolutionary learning of dynamic probabilistic models with large time lags

In this paper, we explore the automatic explanation of multivariate time se ries (MTS) through learning dynamic Bayesian networks (DBNs). We have devel oped an evolutionary algorithm which exploits certain characteristics of MT S in order to generate good networks as quickly as possible. We compare thi s algorithm to other standard learning algorithms that have traditionally b een used for static Bayesian networks but are adapted for DBNs in this pape r. These are extensively tested on both synthetic and real-world MTS for va rious aspects of efficiency and accuracy. By proposing a simple representat ion scheme, an efficient learning methodology, and several useful heuristic s, we have found that the proposed method is more efficient for learning DB Ns from MTS with large time lags, especially in time-demanding situations. (C) 2001 John Wiley & Sons, Inc.