Predicting Cumulative Number of Failures in Software Using an ANN-PSO Based Approach

Software project managers need information such as cumulative number of failures present in a software after testing a certain period of time to determine release time of software. In this paper, an artificial neural network (ANN) based model which uses a new network architecture is proposed to predict cumulative number of failures in software. An extra layer is added between input layer and hidden layer of ANN which uses logarithmic activation function to scale the inputs of ANN. An ANN-PSO based approach is developed in which Particle Swarm Optimization (PSO) method is used to train the ANN. The experiment is carried out using three data sets available in literature and results are compared with existing models found in literature. The results shown that the proposed method is able to produce better prediction than some existing models.