Degree

Doctor of Philosophy (PhD)

Department

Division of Computer Science and Engineering

Document Type

Dissertation

Abstract

Requirements traceability provides support for important software engineering activities. Requirements traceability recovery (RTR) is becoming increasingly important due to the numerous benefits to the overall quality of software. Improving the RTR problem has become an active topic of research for software engineers; researchers have proposed a number of approaches for improving and automating RTR across the requirements and the source code of the system. Textual analysis and Information Retrieval (IR) techniques have been applied to the RTR problem for many years; however, most of the existing IR-based methodologies applied to the RTR problem are semiautomatic or time-consuming, even though many links are correctly recovered using IR. Thus, there is a need for effective and innovative approaches for automatization in the RTR problem. In this research, we study IR techniques applied to the RTR problem to determine the optimal alternative to RTR across the textual content of requirements and system source code, and propose innovative methodologies based on computational intelligence combine with IR to achieve automatization. We approach the study of the RTR problem as an optimization problem; the problem is formulated as a multi or mono objective search in which we assign one-to-many relationships between each requirement and source code classes by considering similarity in their textual content. The Non-dominated Sorting Genetic Algorithm (NSGA-II) and Artificial Bee Colony (ABC), when combined with IR techniques, appear to provide promising alternatives for finding a complete and accurate list of traceability links. We adapt the NGSA-II and ABC algorithms to solve the RTR problem, generate programing tools for experimentation, and report the results on three open source projects. Results show values of precision and recall above 70%. NSGA-II and ABC are also analyzed based on time complexity using the big-O notation; results indicate NSGA-II is more time efficient and less precise than ABC.

Date

4-27-2020

Committee Chair

Carver, Doris

Available for download on Friday, July 09, 2021

Share

COinS