Software Fault Prediction Using Particle Swarm Optimization and Random Forest

Software fault prediction deals with the identification of software faults in the early phases of the software development cycle. It is essential to detect the defects in software as early as possible for the smooth functioning of the software and to reduce the resources and time taken for its maintenance in the future. It can be done either manually, or by using automatic predictors. As the complexity of the software increases, it becomes hard to identify the faults manually. So, to deal with the faults in a timely and more accurate manner, there are many automatic predictors already in use, and various new ones are also being proposed by several researchers. In this study, a method to improve the fault prediction rate in software fault prediction is proposed by combining particle swarm optimization (PSO) with the random forest (RF) classifier. NASA MDP Datasets, which are considered large-scale datasets, are utilized to test the proposed model. The findings show that PSO with RF classifier increases performance when applied on the NASA MDP Datasets and overcomes prior research constraints.