Genetic algorithm-based multi-objective optimization model for software bugs prediction

Abstract

The accuracy and reliability of software are critical factors for consideration in the operation of any electronic or computing device. Although, there exist several conventional methods of software bugs prediction which depend solely on static code metrics without syntactic structures or semantic information of programs which are more appropriate for developing accurate predictive models. In this paper, software bugs are predicted using a Genetic Algorithm (GA)-based multi-objective optimization model implemented in MATLAB on the National Aeronautics and Space Administration (NASA) dataset comprising thirty-eight distinct factors reduced to six (6) major factors via the use of the Principal Component Analysis (PCA) algorithm with SPSS, after which a linear regression equation was derived. The developed GA- based multi-objective optimization model was well-tried and tested. The accuracy and sensitivity level were also analyzed for successful bug detection. The results for optimal values ranging from 95% to 97% were recorded at an average accuracy of 96.4% derived through MATLAB-implemented measures of critical similarities. The research findings reveal that the model hereto proposed will provide an effective solution to the problem of predicting buggy software in general circulation.