Multi-Objective Optimization for Ensemble of Models in Software Reliability Prediction

Software reliability models are very useful to estimate the probability of thesoftware fail along the time. Several different models have been proposed topredict the software reliability growth models (SRGM) however; none of themhas proven to perform well considering different project characteristics. Thevariability of predictive accuracy seems mainly due to the unrealisticassumptions in each model, there is no single model yet available has beenshown to be sufficiently trustworthy in most or all applications. GeneticAlgorithms can proposed the solution by overcome the uncertainties in themodeling. This is dependent on the successful software runs by combiningmultiple models using multiple objective function to achieve the bestgeneralization performance where. The objectives are conflicting and nodesign exists which can be considered best with respect to all objectives. Inthis paper, experiments were conducted to confirm these hypotheses. Thenevaluating the predictive capability of the ensemble of models optimized usingmulti-objective GA has been calculated. Finally, the results were comparedwith traditional models.