Performance Evaluation of Software Development Effort Estimation Using Neuro-Fuzzy Model

ne of the greatest challenges for software developers is forecasting the development effort for a software system for the last decades. The capability to provide a good estimation on software development efforts is necessitated by the project managers. Software effort estimation models divided into two main categories: algorithmic and non-algorithmic. Developers should be able to achieve practices containing effort estimation based on their own programs. New paradigms as Fuzzy Logic may offer an alternative for software effort estimation. In this paper the estimated time for Neuro fuzzy model created for three membership functions is compared with the existing neural network models. The Neuro fuzzy model for Gaussian, triangular and trapezoidal membership function is compared with the neural network models. For this experiment Lopez Martin dataset was used with 41 modules. We compared the three different membership function models (Gaussian MF, Triangular MF and Trapezoidal MF) with the existing neural network models on the basis of five different parameters. Those are Magnitude Relative Error (MRE), Mean Magnitude Relative Error (MMRE), Prediction (Pred), Balanced Relative Error (BRE), Relative standard deviation (RSD) and Root Mean Squared Error (RMSE). Finally it is observed from the comparison that Neuro Fuzzy model using Trapezoidal membership function gives better results than all other models. It is also observed that Trapezoidal MF gives better results for all the five parameters. Keywords—Software development Effort Estimation, MMRE, Pred, BRE, RSD, RMSE, GMF, Tri MF and Trap MF,