Software effort prediction with algorithm based frameworks

Abstract

The need for accurate effort predictions for projects is one of the most critical and complex issues in the software industry. Effort estimation is an essential activity in planning and monitoring software project development to deliver the product on time and within budget. The competitiveness of software organizations depends on their ability to predict the effort required for developing software systems accurately. Several algorithmic approaches have been proposed in the literature to support software engineers in improving the accuracy of their estimations. The decision of how to select different techniques considering the characteristics of a specific dataset through genetic algorithms could be viewed as a search-based problem for software engineering. Some groups use well known evolutionary learning algorithms (MMRE, PRED (25), PRED (50) and PRED (75), while others use machine learning methods like artificial neural network and fuzzy logic. Analogy estimation is also being utilized for estimating software effort, which has been shown to predict accurate results consistently. Genetic frameworks are becoming famous for selecting best learning schemes for effort prediction using the elitism techniques, utilizing some well-known parameters like Spearman's rank correlation, the median of the magnitude of relative error (MdMRE), and mean of the absolute residuals (MMAR). The technique of using fuzzy rough sets for the analysis is also becoming popular. This article aims to compare various algorithm based frameworks for predicting software development effort which can generate the best-optimized software product.