ABSTRACT Software cost Estimation is the process of predicting the amount of time (Effort) required to build a software system. The primary reason for cost estimation is to enable the client or the developer to perform a cost-benefit analysis. Effort Estimations are determined in terms of person-months, which can be translated into actual dollar cost. The accuracy of the estimate will be depending on the amount of accurate information of the final product. Specification with uncertainty represents a range of possible final products, and not one precisely defined product. The input for the effort estimation is size of the project and cost driver parameters. A number of models have been proposed to construct a relation between software size and Effort but no model consistently and effectively predict the Effort. Accurate software effort estimation is a challenge in the software Industry. In this paper a Particle Swarm Optimization technique is proposed which operates on data sets which are clustered using the K-means clustering algorithm. PSO has been employed to generate parameters of the COCOMO model for each cluster of data values. The clusters and effort parameters are then trained to a Neural Network by using Back propagation technique, for classification of data. Testing of this model has been carried out on the COCOMO 81 dataset and also the results have been compared with standard COCOMO model and as well as the neuro fuzzy model. It is concluded from the results that the neural networks with efficient tuning of parameters by PSO operating on clusters, can generate better results and hence it can function efficiently on ever larger data sets.
