Feasibility Study of Machine Learning & AI Algorithms for Classifying Software Requirements
Ullal Akshatha Nayak, K. S. Swarnalatha, A Balachandra · 2022 IEEE 2nd Mysore Sub Section International Conference (MysuruCon) · 2022
Software requirements[15] description and classification is the fundamental and most important activity in the software engineering process. Requirements are obtained through an elicitation process which generally involves interaction with stakeholders such as; exchange of information in person, on notes, by email, on phone, through meetings, etc., which involves a communication language such as English. The description of requirements (ex: functional, non-functional, related others) encompasses few properties such as; understandability, completeness, accuracy, clarity, unambiguousness, testability and related others. Classifying requirements into functional and non-functional category using Machine learning approaches have proved to be successful in the past. The goodness of software requirement properties impact’s the quality levels during the development of a software product and on the resulting product quality. The classification should address semantic details and implicit information during classification to completely satisfy a requirement. This paper presents results of applying different ML algorithms using a simple problem (and data set) for classifying software requirements. The requirements have been described in English following semantic language rules adopted to ease the writing process. The requirement may be obtained from a use case tool (for example rational unified software) or alternate sources. The purpose of this research work is for understanding the application and use of Machine Learning algorithms for the problem of requirements classification, while providing inputs for developing a “software requirements definition and description framework” using English language.