A Systematic Approach of Dataset Definition for a Supervised Machine Learning Using NFR Framework

Non-functional requirements describe important constraints upon the software development and should therefore be considered and specified as early as possible during the system analysis. Effective elicitation of requirements is arguably among the most important of the resulting recommended RE practices. Recent research has shown that artificial intelligence techniques such as Machine Learning and Text Mining perform the automatic extraction and classification of quality attributes from text documents with relevant results. This paper aims to define a systematic process of dataset generation through NFR Framework catalogues improving the NFR's classification process using Machine Learning techniques. A well-known dataset (Promise) was used to evaluate the precision of our approach reaching interesting results. Regarding to security and performance we obtained a precision and recall ranging between ~85% and ~98%. And we achievement a F1 above ~79% when classified the security, performance and usability together.