Identification and classification of functional and non-functional software requirements using machine learning.

In software engineering, it is now necessary to divide needs into functional and non-functional categories to ensure proper operation and overall performance. This study aims to distinguish between functional (F) and non-functional (NF) requirements for application software and classify them accordingly. A portion of the data has been obtained from online resources. In this study, data is cleaned using normalization steps and then proceeded with subsequent steps such as text preprocessing and vectorization. Bag of Words, Term Frequency-Inverse Document, Featurization and Machine Learning Models, ROC and AUC curves, Bi-Grams and n-Grams in Python, Word2Vec, and confusion matrix were covered. Early detection of NFRs allows us to make initial design decisions. Machine Learning algorithms were applied to identify and classify functional and non-functional requirements for application software development based on user requirements. This paper aims to find out which ML algorithm and which model of NLP gives better accuracy and best performance for binary classification. In this study, the combination of BoW and Multinomial Naive Bayes provides the best performance for binary classification. This paper classifies functional Requirement (FR) and non-functional requirements (NFRs) and assists the software developer, software designer, testers, etc., in developing applications. This paper is also helpful for faster software development and delivery. This step facilitates the creation of an SRS document and avoids unnecessary work and complexity during the requirement analysis phase. [ABSTRACT FROM AUTHOR]