Statistical simulation and prediction in software reliability
Alexei Leahu, Carmen Elena Lupu · 2008
On the base of statistical simulation (Monte Carlo method), in this paper it was investigated the rate (relative frequencies) of ”success” in predicting the number of initial (remained) errors by means of Maximum Likelihood Principle. Some numerical results it will be discussed. 1. Model’s description As an extension of the paper [1], the aim of this paper is to use the statistical simulation for some Jelinski-Moranda’s software reliability models in order to check the efficiency of the well known maximum likelihood statistical estimators for some parameters. More exactly, we consider the Jelinski-Moranda (JM) models based on the following hypotheses: 1. The total number N of errors existing initially in the software is unknown constant. 2. Each error is eliminated with probability p =1 , independently of the past trials, repair of the error being snapshot and without introduction of the new errors or 2 � .Each error is eliminated with probability p, 0 <p< 1, independently of the past trials, repair of the error being snapshot and without introduction of the new errors.