Genetic Algorithm for Optimizing Neural Network Based Software Cost Estimation

Software engineering cost models and estimation techniques are used for number of purposes. These include budgeting, tradeoff and risk analysis, project planning and control, software improvement and investment analysis. The proposed work uses neural network based estimation, which is essentially a machine learning approach, is one of the most popular techniques. In this paper the author has proposed a 2 step process for software effort prediction. In first phase known as training phase neural network selects the matching class (datasets) for the given input, which is improved by optimizing the parameters of each individual dataset by Genetic algorithm. In second step known as testing phase, the prediction process is done by adaptive neural networks. The proposed method uses COCOMO-II as base model. The experimental results show that our method could significantly improve prediction accuracy of conventional Artificial Neural Networks (ANN) and has potential to become an effective method for software cost estimation.