Software Reliability Prediction using Neural Network with Encoded Input

摘要: neural network based software reliability model to predict the cumulative number of failures on Feed Forward architecture is proposed in this paper. Depending upon available failure count data, execution time encoded using Exponential and Logarithmic function order provide value as input network. The effect encoding different parameter prediction accuracy have been studied. terms hidden nodes has also performance approach tested eighteen data sets. Numerical results show that giving acceptable across projects. compared with some statistical models change point considering three datasets. comparison a good capability. Software engineers can measure or forecast models. In real development process, cost are limited. used develop reliable under given constraints. managers determine release help these past few years, large models/statistical developed. Every certain assumptions. So predictive capability for There exist no single best suit all cases. influenced by external parameters. To overcome problems, non-parametric like support vector machines last years. not any assumptions which unrealistic situation. Neural become an alternative method modeling, evaluation prediction. literature, many their proven better than (4-6). All basically experimental nature. Anyone who wants apply set, they first find out hit trial method. cases, people lose confidence high level confidence. work, scheme build It shown (14) above two But (14), determined doing repeated experiments purpose paper guideline selection gives consistent applied sets found result existing points. rest organized follows: Section 2, works related introduced. 3 presents schemes such encoding. 4 shows result. Finally, 5 concludes