Neueural-Network-based approach on reliability prediction of software in the maintenance phase

Maintenance of software involves debugging of errors and implementations of enhancement requested by users, these both cause the reliability of software decreased. For the systems that have been used for a considerably long period of time, the various details concerning the initial development phase are usually not known to the users who are responsible for the maintenance of these systems. These cause the estimation of software reliability more difficult. In this paper, a prediction model based on back-propagation neural network (BPN) is proposed to estimate the failures of the software system in the maintaining phase. The ¿failure correction¿ records and the ¿enhancement¿ records are chosen as the input data of the prediction model, the future failure time is the output. A numerical example of a commercial shop floor control system (SFC) is used to illustrate the validation and application of the proposed method.