Optimal selection of failure data for reliability estimation based on a standard deviation method

Software reliability is one of software quality factors and it mostly concerns with the failures of the software. Software failures cost money and it affects to the customer satisfaction, as a result the reliability practice is important. However it is evident to say that software reliability practice in the industry is poor. Impairment of accuracy and the complexity of calculations are some of reasons for lack of applicability. Software reliability is estimated using software reliability estimation models based on observed failure data. The software failure behavior change as a result of software code changes during the testing and is a limitation of most reliability estimation models. This fact has not been considered in most of the reliability models. Our study introduces a simple and accurate method for optimal data selection. The optimal failure data are selected based on the fact that the mean is an estimator for sample data and standard deviation is an estimator for variation of sample data from the mean value. The variation of failure data reflects the fact that the software failure behavior is changed during the testing due to the changes that takes place in code. This variation is examined based on the plotted mean values of time to failure and standard deviation values of time to failure against the failure number in the same graph. This novel method for optimal selection of failure data is simple and accurate method.