A new evolutionism based self‐adaptive multi‐objective optimization method to predict software cost estimation. (24th May 2022)

• Record Type: Journal Article
• Authors: Gouda, Sunil Kumar; Mehta, Ashok Kumar
• Is Part Of: Software, practice & experience. Volume 52:Number 8(2022)
• Page Start: 1826
• Page End: 1848
• Publication Date: 2022-05-24
• Subjects: multi‐objective differential evolution -- multi‐objective evolutionary algorithms -- optimization -- software cost estimation; Computer software -- Periodicals; Computer programming -- Periodicals; Computer programs -- Periodicals; 005.3
• DOI: 10.1002/spe.3092
• Languages: English
• ISSNs: 0038-0644

Abstract

Abstract: One of the major challenge that organizations face in the present environment is having an efficient model for software cost estimation (SCE). In this article, the significance of the meta‐heuristic algorithm in addressing various optimization challenges faced in mathematical models and software applications is discussed. The proposed method uses the new evolutionism‐based self‐adaptive mutation operator to solve the multi‐objective optimization problems. This approach addresses the issues that exist in multi‐objective differential evolution algorithms. To improve diversity among candidate solutions, the Pareto optimality principle is integrated with the evolutionism‐based self‐adaptive mutation operator in a multi‐objective DE algorithm. To reduce the time complexity of Pareto dominance, we have adopted the non‐dominated sorting algorithm. We used eight benchmark test functions to evaluate the effectiveness of the proposed method, and it outperformed the most recent multi‐objective evolutionary algorithms (MOEAs). Furthermore, this article explores software engineering problems like SCE by using the proposed approach, where SCEs are accurately predicted by optimizing the tuning parameters of the multi‐objective constructive cost model. The proposed algorithm achieves better cost prediction as compared to the other standard benchmark algorithms for all objective problems in terms of prediction, mean absolute error, and root mean square error.