Using meta-cognitive sequential learning Neuro-fuzzy inference system to estimate software development effort

Abstract

Software development effort estimation plays a crucial role in the successful completion of any software development project. Estimating the software development effort accurately is one of the challenges in the software industry. In this paper a soft computing technique named Meta Cognitive sequential learning is used to improve the software development effort estimate of a software project. Different datasets namely, Deshernais, NASA, Kitchenham, Maxwell, Telecom, IBM, Kemerer, Hallmark and ISBSG are used in the training and testing of the estimation process. The performance comparison of the former method is compared with other methods such as Particle swarm optimization, Genetic Algorithm and Back propagation network using Mean magnitude of relative error (MMRE) parameter. It is found that the MMRE of Meta cognitive neuro fuzzy method is less than the MMRE of other methods taken for comparison. Thus the Meta cognitive neuro fuzzy method can produce a better estimate of the software development effort and hence can produce an estimate with much precision to the actual effort.