An Application of the Chains-of-Rare-Events Model to Software Development Failure Prediction

Some of the best known models for software reliability are based on non homogeneous Poisson processes. Here, we analyze the application of the Chains-of-Rare-Events model to model grouped failures production. As it has been previously shown, this model can be analyzed as a compound Poisson with a Poisson Truncated at Zero as the compounding distribution. We introduce the mode estimator for the parameter of the Poisson Truncated at Zero. This estimator has the important characteristic of quickly reaching stability around the true value. We apply this model to several data and compare it with a non homogenous Poisson process model, and the Poisson distribution compounded by a geometric model.