Classification of Non-Functional Requirements Using Semi-Supervised Learning Approach

Main Article Content

Devendra Kumar, Anil Kumar, Laxman Singh

Abstract

The primary emphasis is on specialized work involving software engineers. Nevertheless, it is equally imperative to document the deficient features of software design, including aspects such as maintainability, reusability, and reliability. In rapid software development methodologies, such as agile, non-functional requirements (NFRs) are frequently neglected. This neglect results in an increased significance of eliminating non-functional requirements in agile-based software and a heightened emphasis on critical tasks during software migration. Misinterpretation of NFRs can be a significant factor contributing to project failure. A computer network simplifies the implementation of 12 key concepts such as NFR heuristics in the context of failing rules. We propose a semi-written classification system to identify useless needs. In this manner, the initial distribution for the NFR is learned using the small set determined during the heuristic phase. This iterative process helps identify additional needs. The aim was to incorporate this approach into individual recommendations to assist analysts and software designers in the architectural process. Using a semi-supervised learning method, NFRs can be effectively identified and classified. In addition, using other information provided by well-written and informal rules allows classification using fewer preexisting methods. The learning process improves the classification performance by leveraging user feedback during training. Our partial maintenance strategy provides over 70% accuracy compared with traditional maintenance using pre-recorded data. This demonstrates the superiority of the semi supervised approach. The researchers also noted that partial care systems require less human effort to enroll than full care and could be further developed according to the guidelines. Currently, our method is better than other distribution tracking methods, and we believe it will improve the performance of collaboration with input from analyst participants.

Article Details

Section
Articles