Software development efforts prediction using artificial neural network

Software project managers need an accurate assessment of software development efforts to achieve reliable software within development budget and schedule. A single layer neural network (SLP) is reported to predict software development efforts from software quality metrics. Particle swarm optimisation for training, principal component analysis (PCA) for dimension reduction of input features and genetic algorithm for optimising artificial neural network architecture are used. Literature reported datasets are tested and the results are acceptable within the limits. However, SLP_NN without pre-processing with PCA is adequate and in some cases, reduction approach may be dropped.