Use of Non-functional Requirements in Software Effort Estimation: Systematic Review and Experimental Results

Abstract

Software effort estimation is a critical step in software project management, and its main driver are requirements. Some algorithmic methods use as inputs only functional requirements and others take also into account non-functional requirements. The goal of this study is to understand the correlation of using non-functional requirements on the accuracy of software effort estimation algorithmic methods. A systematic literature review was conducted to learn which non-functional requirements are used, how they are used, and their effects on estimation accuracy. The systematic review shows that only 33% of 39 algorithmic methods use non-functional requirements. However, the investigation on its correlation with estimation accuracy was not conclusive from published results. In order to address this issue, a quasi-experiment was conducted on publicly available datasets. This experiment shows that the use of non-functional requirements results in a reduction of about 30% in the estimation error, with statistically significant confidence.