Neural Net Back Propagation and Software Effort Estimation: A comparison based perspective

In order to achieve accurate estimates a number of contributors proposed and validated several algorithmic estimation techniques to eradicate or reduce the issue of misleading estimates. However, these least square regression based algorithmic techniques are considered vulnerable while dealing with complex non-linearity in variables. In this article we present the main findings of few research papers that have utilized a non-linear approach, neural networks back propagation (in some cases amalgamated with other computational intelligence technique), in software estimation for estimation purposes. The selection of these research papers was similarity among them based on the use of back propagation neural net for prediction of effort and comparisons among the results produced by statistical approaches like Magnitude of Relative Error, Mean Magnitude of Relative Error, Coefficient of determination and pred(l).