RESUMEN

Researchers used four software reliability growth models in the final test phases of three embedded software projects to predict the software's remaining faults. They compared the models' predictions to expert predictions and the actual results. The models outperformed the experts in predicting the total number of faults at 25 percent of the elapsed test-time. The researchers concluded that software reliability growth models are useful for supporting management decisions during a software product's final test phases, provided they're combined with results from other estimation methods.