Product Requirement Information Using TF-IDF and Naive Bayes Techniques as a Remedy for Stakeholders in IT-Enterprises
Serda Hauser · 2022 2nd International Conference on Consumer Electronics and Computer Engineering (ICCECE) · 2022
Writing adequate requirements necessitates a delicate sensibility on the researcher's part. It is not uncommon for requirements to go through numerous iterations before all stakeholders understand them. In this research study, unspecified requirements are detected as incomplete written functional and non-functional requirements and analyzed using a specification matrix. The matrix presents a new metric for assessing whether requirements are fully specified, moderately specified, or unspecified. The specification matrix is put through a UseCase within Deutsche Telekom (DT) AG. The unspecified requirements are subjected to a semantic review using a requirements stencil, referred to as adequate product requirement information (PRI). The adequate PRI are automatically classified by categories, including chapters, using statistical approaches (TF-IDF and Naive Bayes probability calculation). To avoid the propagation of unspecified requirements, this work seeks to evaluate stakeholder-oriented novel solutions utilizing the specification matrix and the TF-IDF with the Naive Bayes probability computation.