An evolutionary ensemble analogy‐based software effort estimation

• Published in: Software, practice & experience Vol. 52; no. 4; pp. 929 - 946
• Main Authors: Shahpar, Zahra, Bardsiri, Vahid Khatibi, Bardsiri, Amid Khatibi
• Format: Journal Article
• Language: English
• Published: Hoboken, USA John Wiley & Sons, Inc 01.04.2022; Wiley Subscription Services, Inc
• Subjects: analogy‐based estimation; Datasets; ensemble effort estimation; Estimation; Evolutionary algorithms; genetic algorithm; Software development; software effort estimation
• ISSN: 0038-0644; 1097-024X
• DOI: 10.1002/spe.3040

Analogy‐based estimation (ABE) is the most commonly applied method for estimating using software. Although there are several analogy‐based estimation techniques, there is no consistent conclusion of which technique is the best in all circumstances. Therefore, this article presents an evolutionary ensemble ABE (EEABE) for software cost estimation. Ensemble effort estimation (EEE) models forecast software development endeavors through using multiple estimation methods. In this article, EEABE combines GA as an evolutionary algorithm with six ABE models. The proposed method has been evaluated on four well‐known datasets comprising Maxwell, Albrecht, Kemerer, and Desharnais, by the k‐fold cross‐validation technique and based on the performance criteria of MRE, MMRE, MDMRE, BMMRE, and PRED(0.25). The simulation results demonstrate that the use of EEABE increases the accuracy of estimation and reduces the cost. Besides, EEABE is a flexible and adaptable model with any type of dataset and software development project, and it has been optimized with the possibility of using various ABE techniques.