Early Prediction of Software Reliability: A Case Study with a Nuclear Power Plant System

Abstract

Existing methods to predict software reliability using the Markov chain are based on assumed state-transition probabilities. A new prediction approach applied to a nuclear plant's feed-water system yielded results that were 96.9 percent accurate relative to the system's actual reliability. Across 38 operational datasets, the average accuracy was 99.67 percent. © 2016 IEEE.