An analysis method for tagged-task models
Hiroei Imai, Masahiro Tsunoyama, Ikuo Ishii, Hideo Makino · Systems and Computers in Japan · 1998
This paper presents a method for obtaining the probabilities of absorbing states in a Markov chain for a tagged-task model having a small number of absorbing states and a large number of transient states. In the method, the states are first divided into cosets based on reachability between states of the chain; then the cosets are divided into subsets recursively to reduce the memory size required for computation. During the computation of probabilities, cosets are aggregated to a representative when the state probabilities are smaller than a given threshold value in order to reduce computation time. This paper also gives an example of the analysis. Errors, memory size, and computation time for the method are compared with those for the corresponding method without any divisions and aggregations. © 1998 Scripta Technica, Syst Comp Jpn, 29(7): 41–49, 1998