Reliable machine learning models for estimating effective software development efforts: A comparative analysis

Abstract

Effort estimation in the software industry is recognised as one of the most critical activities for the success of the overall solution delivery in the software development process. Software Development Effort Estimation (SDEE) predicts the efforts to develop any software project in terms of persons-month or person-hours. Effort prediction in the early stages of the software development life cycle (SDLC) is always been a challenge. The primary objective of this paper is to analyse different machine learning models and to identify a stable and accurate estimation model for efficient effort estimation. The paper performs a comparative analysis, to analyze the performance of eight different machine learning models, with commonly used datasets in the domain of Software Development Effort Estimation, in terms of eight metrics. On performing experiments and comparing metrics, the analysis showed that the various machine learning estimation models outperformed some particular datasets, hence it can be concluded that machine learning boost-up the performance in the domain of SDEE. Moreover, to conclude, the paper also compared the models based upon the first three choices of models and identified Random Forest has comprehensive superiority and stability in terms of accuracy in estimation, followed by other models.