关键词: 软件成本估算, RBF神经网络, 遗传算法, 样本聚类

Abstract: Accurate software development cost estimation is very important for effective project management.So far，no model has been proved to be successful for predicting software development cost.As the accuracy of estimating software cost is not satisfactory，this study is designed to find a new more precise method.A novel radial basis function neural network is proposed to estimate the software cost.Hidden layer nodes are decided by sample cluster.Genetic Algorithm and the linear least squares algorithm are used as the learning method for this model.The experimental results show that model can estimate software cost accurately.

Key words: software cost estimation, RBF neural network, Genetic Algorithm, sample cluster