To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
A Novel Method for Software Effort Estimation Using by MLPNN, Meta-Heuristic Algorithms and Regression Based Methods |
|
Author |
Mahdi Khazaiepoor, Amid Khatibi Bardsiri |
| Citation |
Vol. 26 No. 1 pp. 145-157 |
|
Abstract |
Today, effort estimation of software development has got a great importance in managing the projects properly. Proper and accurate cost estimation helps the customers and investing and also has a determinate role in logical decision making during doing and managing a project. Various models of cost estimation has been invented and used up to now. These models try to estimate the amount of necessary effort for doing a project. Most of the provided models act based on the data collected from the previous projects. Some of the efficient instruments for making these models are regression, neural network and meta-heuristic algorithms. In this paper we have used Multiple Layer Perceptron (MLP) neural network and Multi Linear Regression (MLR) and combined them with Genetic Algorithm (GA) and Imperialist Competition Algorithm (ICA). As a result we reached a model for effort estimation that can do the software projects effort estimation with less inaccuracy comparing with the previous methods. For this purpose, the data collections of COCOMO, Maxwell and Albrecht have been used which are standard and available for assessment and comparing the suggested model. Due to the given result, average performance improvement of suggested model for MMRE performance criterion on each data collections is 23%, 38% and 35%. |
|
Keywords |
Effort Estimation of software development, Multi Linear Regression (MLR), neural network, Genetic Algorithm (GA), Imperialist Competition Algorithm (ICA), Maxwell, Albrecht, COCOMO. |
|
URL |
|
|
Title |
A Novel Method for Software Effort Estimation Using by MLPNN, Meta-Heuristic Algorithms and Regression Based Methods |
|
Author |
Mahdi Khazaiepoor, Amid Khatibi Bardsiri |
| Citation |
Vol. 26 No. 1 pp. 145-157 |
|
Abstract |
Today, effort estimation of software development has got a great importance in managing the projects properly. Proper and accurate cost estimation helps the customers and investing and also has a determinate role in logical decision making during doing and managing a project. Various models of cost estimation has been invented and used up to now. These models try to estimate the amount of necessary effort for doing a project. Most of the provided models act based on the data collected from the previous projects. Some of the efficient instruments for making these models are regression, neural network and meta-heuristic algorithms. In this paper we have used Multiple Layer Perceptron (MLP) neural network and Multi Linear Regression (MLR) and combined them with Genetic Algorithm (GA) and Imperialist Competition Algorithm (ICA). As a result we reached a model for effort estimation that can do the software projects effort estimation with less inaccuracy comparing with the previous methods. For this purpose, the data collections of COCOMO, Maxwell and Albrecht have been used which are standard and available for assessment and comparing the suggested model. Due to the given result, average performance improvement of suggested model for MMRE performance criterion on each data collections is 23%, 38% and 35%. |
|
Keywords |
Effort Estimation of software development, Multi Linear Regression (MLR), neural network, Genetic Algorithm (GA), Imperialist Competition Algorithm (ICA), Maxwell, Albrecht, COCOMO. |
|
URL |
