On parameter estimation of software reliability models

ABSTRACT

Most software reliability models use the maximum likelihood method to estimate the parameters of the model. The maximum likelihood method assumes that the inter-failure time distributions contribute equally to the likelihood function. Since software reliability is expected to exhibit growth, a weighted likelihood function that gives higher weights to latter inter-failure times compared to earlier ones is suggested. The accuracy of the predictions obtained using the weighted likelihood method is compared with the predictions obtained when the parameters are estimated by the maximum likelihood method on three real datasets. A simulation study is also conducted.

KEYWORDS:

• Non-homogeneous Poisson process models
• Software reliability growth
• Weighted likelihood method

MATHEMATICS SUBJECT CLASSIFICATION:

• 62 Statistics
• 68 M15
• Reliability

Acknowledgments

The author is grateful to the Editor-in-Chief and to the anonymous referee(s) for their useful suggestions that improved an earlier version of this paper.