The determination of optimal software release times at different confidence levels with consideration of learning effects

Abstract

As most software reliability models do not clearly explain the variance in the mean value function of cumulative software errors, they might not be effective in deducing the confidence interval regarding the mean value function. In such cases, software developers cannot estimate the possible risk variation in software reliability by using the randomness of the mean value function, thus reducing the decision‐making reliability when determining an optimal software release time. In this paper, the method of stochastic differential equations is used to build a software reliability model, which is validated based on practical data previously used in six published papers. Moreover, the estimation of the parameters of the proposed model, which can be defined as the autonomous error‐detected factor and the learning factor, is also illustrated, and the results of model validation empirically confirm that the proposed model is able to account for a fairly large portion of the variance of the mean value function. Additionally, the confidence intervals of the mean value function regarding software faults are employed to assist software developers in determining the optimal release times at different confidence levels. Finally, a numerical example is given to verify the effectiveness of the proposed model. Copyright © 2008 John Wiley & Sons, Ltd.

Debugging‐workflow‐aware software reliability growth analysis

“…Non‐homogenous Markov chains were also proposed to model these two processes in a unified way . The incorporation of the debuggers' learning effect into SRGM modelling has been explored by Ho et al with stochastic differential equations for the correction process. Queuing models for nonimmediate fault corrections have been proposed by Musa et al and by Huang and Huang .…”

Neural network for software reliability analysis of dynamically weighted NHPP growth models with imperfect debugging

“…Roy et al proposed an NHPP‐based S‐shaped Software Reliability Growth Model (SRGM) in presence of imperfect debugging with a new exponentially increasing fault content function. The parameter estimation through the autonomous error‐detected factor was explained by Wen . NHPP‐based SRGMs have been attempted in classification of software reliability into different classes, such as modifying Goel and Okumoto , NHPP , S‐shaped growth , discrete reliability growth , multivariable model , and generalized NHPP .…”

Risk-based software release policy under parameter uncertainty

“…The software testing process is very costly and a delayed release of software can result in a significant loss of market share and high levels customer dissatisfaction. Because of this, when to release the software is of great practical concern and this is a widely discussed problem [7][8][9][10][11][12][13][14][15][16][17]. Generally, the optimal release policy is investigated based on two criteria: either meeting the reliability requirement or minimizing the total expected cost.…”

“…Generally, the optimal release policy is investigated based on two criteria: either meeting the reliability requirement or minimizing the total expected cost. Recently, the uncertainty involved in the determination of the optimal release time has received close attention [15,16]. It has been pointed out that the point estimate produced using currently available methods is not precise in that the software debugging process is essentially random.…”

“…*Corresponding author: Department of Industrial and Systems Engineering, National University of Singapore, Kent Ridge, Singapore 119260, Singapore. email: mxie@nus.edu.sg Recently, the uncertainty involved in the determination of the optimal release time has received close attention [15,16]. It has been pointed out that the point estimate produced using currently available methods is not precise in that the software debugging process is essentially random.…”