Research Article
Identification of Nonfunctional Requirement Conflicts: Machine Learning Approach
@INPROCEEDINGS{10.1007/978-3-030-93709-6_29, author={Getasew Abeba and Esubalew Alemneh}, title={Identification of Nonfunctional Requirement Conflicts: Machine Learning Approach}, proceedings={Advances of Science and Technology. 9th EAI International Conference, ICAST 2021, Hybrid Event, Bahir Dar, Ethiopia, August 27--29, 2021, Proceedings, Part I}, proceedings_a={ICAST}, year={2022}, month={1}, keywords={Nonfunctional requirements Requirement conflict Machine learning Conflict catalog Natural language processing (NLP) Pre-trained so word2vec embedding}, doi={10.1007/978-3-030-93709-6_29} }- Getasew Abeba
Esubalew Alemneh
Year: 2022
Identification of Nonfunctional Requirement Conflicts: Machine Learning Approach
ICAST
Springer
DOI: 10.1007/978-3-030-93709-6_29
Abstract
The most common causes of software failure in system development are requirements issues. One of these issues is requirement conflicts, which results in expensive costs and a long development time. This is because contradicting requirements make it difficult to design, test, and maintain a software system, which almost always results in software failure. Using manual and semi-automated methods, many researchers attempted to overcome the challenge of detecting conflicting requirements. We’ve suggested a machine learning-based model for detecting conflicts between non-functional requirements in a Software Requirement Specification (SRS) document. To build the model for identifying non-functional requirement conflicts; text preprocessing, vectorization, and classification are included. The text from the document is preprocessed into a series of words using natural language processing (NLP). Then, using vectorization techniques to give words weight, a series of words are stored in numeric representation and utilized as input for classification algorithms. The prepared dataset is used to test traditional machine learning and deep learning classification techniques. Bi-LSTM with pre-trained SO word2vec embedding performs 84.74% accurately, according to a comparative experimental investigation. Future research directions in the problem domain include identifying the relationship between quality attributes and resolving nonfunctional requirements conflict through experiments.
