About 50% of the total cost of developing software is the cost of its testing and debugging. Software Reliability Growth models—which play an important role in planning and managing testing activity—are found to be appropriate in determining when the error levels are low enough to stop debugging. Although many Software Reliability Growth Models (SRGMS) have been proposed as continuous time models to plan and execute testing, there is a need to evolve discrete models to cater to discrete data. SRGMS have evolved over the years from simple to flexible models. This paper proposes the combination of two simple fault-based models to create a dynamically weighted model which describes both concave and S-shaped growth of mean value functions, occurring either in the whole project or in parts thereof. Most importantly, this combination model can be successfully deployed without modification for testing all the releases of a software product. On validation with a number of discrete datasets and releases, this model is found to provide satisfactory goodness of fit. In addition to being adequate for past failure data, predictive validity measures have confirmed that the proposed model will be equally adequate for future data also.

开发软件的总成本中约有50％是测试和调试的成本。在确定错误级别何时足够低以停止调试时，软件可靠性增长模型（在计划和管理测试活动中起着重要作用）非常合适。尽管已经提出了许多软件可靠性增长模型（SRGMS）作为计划和执行测试的连续时间模型，但是仍需要发展离散模型以适应离散数据。多年来，SRGMS已从简单模型演变为灵活模型。本文提出了两个简单的基于故障的模型的组合，以创建一个动态加权模型，该模型描述了在整个项目或部分项目中均值函数的凹形和S形增长。最重要的是，这种组合模型可以成功部署，而无需进行修改即可测试软件产品的所有发行版。经过大量离散数据集和版本的验证，发现该模型具有令人满意的拟合优度。除了适用于过去的故障数据之外，预测有效性的措施已经证实，提出的模型也同样适用于将来的数据。