Bayesian Network Based Software Reliability Prediction by Dynamic Simulation

In order to solve the problem faced by reliability prediction and analysis for large-scale complex software system, Bayesian Network based software reliability modeling method and task flow oriented software reliability simulation prediction method are proposed in this paper. Bayesian Network based reliability modeling can calculate the initial reliability for complex software system by structure learning and parameter learning from the software architecture and the possible history data, on the basis of which Monte Carlo simulation can be used to setup the reliability logical relationship between different tasks in software system to realize the dynamic reliability prediction. This proposed method can comprehensively utilize the priori information of software architecture, history data and software task flows to conduct the dynamic reliability prediction and find the reliability weaknesses at the same time. One Train network Control & Management System (TCMS) software is selected as the experiment application to verify its feasibility and validity.