The director of the Higher Polytechnic School and researcher at the Universidad Europea del Atlántico (European University of the Atlantic, UNEATLANTICO), Dr. Manuel Masías Vergara, and UNEATLANTICO researcher, Ernesto Bautista Thompson, collaborate in a study that analyzes and compares the evolution of cost and effort estimation techniques in software projects.
Despite the fact that several research have been conducted on software cost and effort estimation, cost and effort overrun problems continue to occur in several projects in the software industry. Erroneous estimates generate economic and time losses. This is evidence of the need for research to evaluate the estimation techniques that have evolved over the years.
In machine learning (ML) and non-ML techniques, for example, accuracy is a frequent problem. Especially when working on complex projects or projects with changing requirements. In order to complete projects on time and within budget, accurate estimation plays an important role, so organizations invest heavily in this factor to ensure a successful project and customer satisfaction. In this context, it is essential to determine accurate estimation techniques to avoid cost and effort overruns.
The objective of this study has been to provide an overview of the current state of cost estimation research and to identify opportunities for future researchers to optimize the accuracy and effectiveness of cost estimation methods. For this purpose, ML and non-ML methods were used to evaluate the challenges and limitations of these methods, such as the indispensable use of historical data and the difficulty of interpreting the results.
The systematic review determined that the most commonly used estimation techniques are ANN and COCOMO, followed by Ensemble and FPA. In addition, ANN has been shown to be more efficient than several ML and non-ML techniques. The most commonly used precision metric is MMRE.
If you want to know more about this fascinating study, click here.
For further research, check the UNEATLANTICO repository.