Aristotle University Of Thessaloniki (AUTH)
Eυφυής μεθοδολογία αποτίμησης του βαθμού διατηρησιμότητας έργων λογισμικού με τη χρήση τεχνικών εξόρυξης δεδομένων
Abstract
dc:descriptionThe modern software development lifecycle requires fast development sprints, incremental integration of features and the ability to change fast. In such a volatile environment, software quality control is of utmost importance. This fact is more than evident considering that numerous software projects fail because they are unable to meet their technical specifications in the context of a constrained timeframe and/or budget. The answer to this difficult challenge of managing technical debt lies in the development of tools and methodologies that enable effective quality assessment, minimize risks and set up a proper mitigation plan. Obviously, evaluating the extent to which a software project adheres to certain quality-related characteristics is an open research question that has drawn the attention of researchers for decades.Given its nature, software quality assessment is a complex and multifaceted problem that requires examining numerous factors varying with respect to the special characteristics of each software project as well as its field of application. This complexity has been identified since the first research efforts in the area of software quality, which highlight the need to disaggregate the high-level concept of software quality into certain entities that provide clarity and broad applicability. To that end and in an effort to create a common ground among the scientific community, several quality standards have been proposed that decompose quality into certain quantifiable characteristics, each related to certain metrics. One of the most important quality-related characteristics is maintainability, which expresses the degree of effectiveness and efficiency with which a product or system can be modified based on the changes in its functional and non-functional requirements. The current doctoral thesis focuses exactly on maintainability and proposes a data- driven methodology for assessing the extent to which a software component is maintainable.The proposed methodology employs data mining techniques for building a hierarchical maintainability evaluation model that harnesses software engineering data while focusing on certain design objectives. The first objective of the doctoral thesis is directed towards providing a fully-automated methodology able to be integrated into the software development process in a continuous manner.Automating the maintainability assessment process is of vital importance given that manually examining the source code, a process widely accepted by the traditional evaluation approaches, is extremely time and effort consuming, especially in large and complex software projects. In this context, the proposed approach utilizes the deluge of data residing in online code hosting facilities and constructs several benchmark datasets that constitute the information basis upon which the constructed maintainability evaluation models are built.Given the principles of object-oriented programming that software artifacts under investigation follow different design objectives, our second design objective refers to constructing a methodology applicable at multiple levels (method, class, and package). This multi-level analysis enables providing a holistic maintainability assessment that covers modelling the behavior of a wide range of software properties along with their progress in time.The remaining objectives of the proposed approach refer to the ability to provide predictions at an early project stage, as well as the provision of interpretable evaluation results. The ability to provide early predictions and identify maintainability-related issues in the early stages of the software development process is critical, as it warns developers for the need of the required refactoring. On the other hand, an interpretable evaluation mechanism provides developers with targeted recommendations on the risks identified as well as the certain code fragments that require auditing.Overall, the proposed maintainability evaluation approach harnesses software engineering data residing in online source code hosting facilities and employs data mining techniques towards constructing various models that enable automated maintainability control. It also enables thorough evaluation using a hierarchical approach that can be easily integrated in the software development process in a continuous and actionable manner.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Papamichail, Michail
- Παπαμιχαήλ, Μιχαήλ
Subjects
dc:subject × 14- Ποιότητα λογισμικού
- Τεχνολογία λογισμικού
- Ανάλυση λογισμικού
- Μετρικές ποιότητας λογισμικού
- Software quality
- Software engineering
- Software analysis
- Software quality metrics
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Υπολογιστές, Υλικό (hardware) και Αρχιτεκτονική
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Computer science, Hardware and Architecture
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/51407