ESTIMATION OF PARAMETERS FOR NONHOMOGENEOUS POISSON PROCESS: SOFTWARE RELIABILITY WITH CHANGE-POINT MODEL

Abstract

Nonhomogeneous Poisson process (NHPP) models, frequently employed in reliability engineering, are used to estimate the number of software errors remaining in a software system. Traditionally, the error detection rate of NHPP models is usually assumed to be a continuous and monotonic function. The error detection rate may, however, not be smooth and can change if the testing environment, strategy or resource allocation is changed. This paper describes NHPP with change-point software reliability models. Due to the irregularity imbedded in the model, the classical maximum likelihood method and the conditional maximum likelihood method cannot be used with interfailure data. The parameters of the NHPP with change-point model can however be estimated by the least squares method. According to the results of a simulation study and analysis of real data, the least squares estimates shown to perform well.

Keywords:

• Software reliability
• Nonhomogeneous Poisson process
• Error detection rate
• Change-point
• Maximum likelihood
• Least squares

ACKNOWLEDGMENTS

The author would like to thank a referee for helpful comments and suggestions. This research was partially supported by NSC 85-2121-M-031-003 from the National Science Council, R.O.C.