PDF

Keywords

Estimation models
COCOMO model
artificial neural network

Abstract

Estimation models in software engineering are used to predict some important and future features for software project such as effort estimation for developing software projects. Failures  of  software  are  mainly  due  to  the faulty  project  management  practices. software project effort estimation is an important step in the process of software management of large projects. Continuous changing in software project makes effort estimation more challenging. The main objective of this paper is find a model to get a more accurate estimation. In this paper we used the Intermediate COCOMO model which is categorized as the best of traditional Techniques in Algorithmic effort estimation methods. also we used an Artificial approaches which is presented in (FFNN,CNN,ENN,RBFN) because of the Ability of ANN(Artificial Neural Network) to  model a  complex  set  of  relationship  between  the  dependent  variable (effort)  and  the  independent  variables  (cost  drivers)which makes  it  as  a  potential  tool    for  estimation. This  paper  presents a  performance analysis of ANNs used in effort estimation. We create and simulate this networks by MATLAB11 NNTool depending on NASA aerospace dataset which contains a features of 60 software project and its actual effort. the result of estimation in  this paper shows that the neural networks in general enhance the performance of traditional COCOMO and we proved that the ENN was the best network between neural networks and the CNN was the next best network and the COCOMO have the worst between the used methods.  
https://doi.org/10.33899/csmj.2013.163464
  PDF