Software release optimization for a non-kalman filter SRGM
Xiangqian Jiang, Donglei Du, Tapabrata Ray, Sean N. Ghazavi · 2007
The optimality of the one-bug-look-ahead (OLA) software release policy proposed in Morali, N. and Soyer, R. (2003) for a non- Kalman filter type of software reliability growth model (SRGM) is disproved in Jiang, X., et al., (2007). The remaining question - what is the optimal policy - is answered in this paper. It turns out that the optimal policy has a control-limit structure similar to OLA, but is not as impatient. A key step in the research is the identification of a Markov structure implied in the Bayesian updating mechanism which in turn reduces the optimal stopping formulation to an easily solvable Markov decision problem. A numerical example based on the NTDS data is included for illustration.