Georgia Southern University
Using Machine Learning to Analyze Software Requirement Dependencies
Abstract
dc:description.abstractAgile methodologies such as Scrum have been increasingly used in software development processes to create high-quality software products more quickly and adaptively. Identifying faults early and analyzing the impact of changing requirements (user stories) is time-consuming and tedious. This research presents a novel approach to supporting the change impact analysis of user stories. The pairwise semantic similarities of the features (TFIDF vectors Word2Vec) extracted from 147 user stories were computed using Cosine similarity and Word Mover's distance measures. After quantifying the similarities of all pairs of user stories, we applied K-Means, Agglomerative, and Spectral clustering to identify semantically similar clusters of user stories. The experimental results of these combinations of techniques obtained using different clustering techniques show varied performance results compared to the ground truth (from the implementation of these user stories). The research findings indicate that the developers can use our approach to locate the inter-related user stories as they fix requirements, which reduces the time otherwise spent on manual analysis.
Degree
thesis:*- Name thesis:degree_name
- Master of Science in Computer Science (M.S.)
- Level thesis:degree_level
- Thesis (restricted to Georgia Southern)
- Discipline thesis:degree_discipline
- Department of Computer Science
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kaur, Mehakpreet
- Contributors dc:contributor
-
- Gursimran S. Walia
- Lixin Li
Subjects
dc:subject × 12Rights
- Licence dc:rights.license
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.georgiasouthern.edu/etd/2424