A New Approach to Software Effort Estimation Using Different Artificial Neural Network Architectures and Taguchi Orthogonal Arrays

Abstract

In this paper, two different architectures of Artificial Neural Networks (ANN) are proposed as an efficient tool for predicting and estimating software effort. Artificial Neural Networks, as a branch of machine learning, are used in estimation because they tend towards fast learning and giving better and more accurate results. The search/optimization embraced here is motivated by the Taguchi method based on Orthogonal Arrays (an extraordinary set of Latin Squares), which demonstrated to be an effective apparatus in a robust design. This paper aims to minimize the magnitude relative error (MRE) in effort estimation by using Taguchi's Orthogonal Arrays, as well as to find the simplest possible architecture of an artificial Neural Network for optimized learning. A descending gradient (GA) criterion has also been introduced to know when to stop performing iterations. Given the importance of estimating software projects, our work aims to cover as many different values of actual efficiency of a wide range of projects as possible by division into clusters and a certain coding method, in addition to the mentioned tools. In this way, the risk of error estimation can be reduced, to increase the rate of completed software projects.

Improved Effort and Cost Estimation Model Using Artificial Neural Networks and Taguchi Method with Different Activation Functions

“…The first proposed ANN architecture in our experiment is ANN-L27. This architecture is based on the Taguchi orthogonal array L27 with 13 parameters [ 1 , 2 , 3 ] (W i , I = 1, …, 13), and three levels: L1, L2, and L3 ( Figure 1 , Table 1 ).…”

“…In recent years, due to a significant evolution in adopting and developing new technologies and methodologies in the area of software effort estimation, many researchers are attempting to optimize the accuracy of this process [1,2]. Since this is one of the crucial processes in completing a software product, there is a continuous need for questioning both overestimation and underestimation [3]. We can relate costs to the project in the initial phase, depending upon the efforts needed [4,5].…”

“…The interval containing the solution is gradually dwindling, as the result of an iterative robust design algorithm. The numerical value of the cost-effect function that is observed in every examination in data distribution is sensitive to individual predictions [ 1 , 3 ]. To increase the stability and safety of the experiment, prediction and correlation between the estimated and actual values are calculated.…”

Extreme Learning Machine Applied to Software Development Effort Estimation

“…Estimating the software development effort is one of the project management processes responsible for determining the effort required to complete the project [2]. Therefore, accurate effort estimation is one of the critical points in reducing risks increasing the chances of success in the projects [3].…”

“…Thereupon, proposing ML techniques for Software Development Effort Estimation (SDEE) is an efficient alternative due to its learning capacity, modifying its behavior autonomously. Furthermore, they assist in decision-making based on data analysis, using minimal human interference (specialist) [3]. Thereby, specialists spend less time estimating the project and more time on other tasks of the system that satisfy the customer [1].…”