An improved particle swarm optimisation-based functional link artificial neural network model for software cost estimation

Abstract

Software cost estimation is the forecast of development effort and time needed to develop a software project. Estimating software cost is endlessly proving to be a difficult problem and thus catches the attention of many researchers. Recently, the usage of meta-heuristic techniques for software cost estimation is increasingly growing. In this paper, we are proposing a technique consisting of functional link artificial neural network model and particle swarm optimisation algorithm as its training algorithm. Functional link artificial neural network is a high order feedforward artificial neural network consisting of an input layer and an output layer. It reduces the computational complexity and has got the fast learning ability. Particle swarm optimisation does optimisation by iteratively improving a candidate solution. The proposed model has been evaluated on promising datasets using magnitude of relative error and its median as a measure of performance index to simply weigh the obtained quality of estimation.