A generalized gamma software reliability model

Abstract

SUMMARYIn this paper, the authors propose a generalized software reliability model based on a nonhomogeneous Poisson process (NHPP), on the assumption that software fault discovery times follow a generalized gamma distribution. The proposed model encompasses representative NHPP models developed before now and is also consistent with probabilistic debugging theory based on the concept of generalized order statistics. Furthermore, in this paper, the authors propose a parameter estimation method which combines the EM (Expectation-Maximization) algorithm and a heuristic solution method, as an effective parameter estimation method for the generalized gamma software reliability model.

Test based safety-critical software reliability estimation using Bayesian method and flow network structure

“…Hence, immediately following equation (11), the reliability of the block composed of the parallelconnected edges under node n i (i.e. the probability of no defect for the edges under the node n i ) is the event that every test has successfully passed.…”

“…R ij = 1. Next, considering the reliability of the combined parallel edges, assuming that the reliabilities of all these edges are ones, using equations (11) and (12), one concludes that the reliability of the combined parallel edges is one. Third, considering the reliability of the combined serial edges whose reliabilities are ones, using equation (18), one concludes that the reliability of the combined serial edges is one.…”

“…Up-to-date, the most extensively investigated software reliability method is the software reliability growth model. 7 Many statistical models, such as the exponential distribution, Poisson distribution, Weibull distribution, and Gamma distribution models have been considered, and many different scenarios have been discussed in previous works [8][9][10][11][12][13] and references therein. The reliability growth model is more like traditional hardware model in that the failure is timerelated.…”

Reliability Estimation for a Digital Instrument and Control System

“…However, the most extensively investigated software reliability method is the software reliability growth model [4]. Many statistics models, such as the exponential distribution, Poisson distribution, Weibell distribution, and Gamma distribution models have been considered, and many different scenarios have been discussed [5][6][7][8]. Nonetheless, there is no formed consensus on an ideal software reliability estimation method.…”