Genetic Algorithm-Based Fuzzy Inference System for Describing Execution Tracing Quality

Abstract

Execution tracing is a tool used in the course of software development and software maintenance to identify the internal routes of execution and state changes while the software operates. Its quality has a high influence on the duration of the analysis required to locate software faults. Nevertheless, execution tracing quality has not been described by a quality model, which is an impediment while measuring software product quality. In addition, such a model needs to consider uncertainty, as the underlying factors involve human analysis and assessment. The goal of this study is to address both issues and to fill the gap by defining a quality model for execution tracing. The data collection was conducted on a defined study population with the inclusion of software professionals to consider their accumulated experiences; moreover, the data were processed by genetic algorithms to identify the linguistic rules of a fuzzy inference system. The linguistic rules constitute a human-interpretable rule set that offers further insights into the problem domain. The study found that the quality properties accuracy, design and implementation have the strongest impact on the quality of execution tracing, while the property legibility is necessary but not completely inevitable. Furthermore, the quality property security shows adverse effects on the quality of execution tracing, but its presence is required to some extent to avoid leaking information and to satisfy legal expectations. The created model is able to describe execution tracing quality appropriately. In future work, the researchers plan to link the constructed quality model to overall software product quality frameworks to consider execution tracing quality with regard to software product quality as a whole. In addition, the simplification of the mathematically complex model is also planned to ensure an easy-to-tailor approach to specific application domains.

Practical Consequences of Quality Views in Assessing Software Quality

“…The abstract notion of software product quality and the complex mathematical computations applied in [67] for reasoning under uncertainty might obscure the practical use of the findings documented in [2,65,67] and discourage potential practitioners. Thus, from the point of view of research methods, the current study contributes to the body of research with (1) a model simplification by approximating the quality model published in [67] with a Generalized Additive Model (GAM) [68], and (2) a practical demonstration of the findings of the authors' previous publications [2,65,67].…”

“…The case studies aim to lead the reader through complex situations full of pitfalls, which might not necessarily appear in each setting. The presentation does not require any familiarity with the uncertainty computations applied in [67], nor a deep understanding of quality modelling in terms of terminology, concepts, and meta-models, which were necessary to prepare the taxonomy and relevance ranking as reported in [2,65].…”

On the Use of Quality Models to Address Distinct Quality Views

“…Galli et al in [5] carried out a comprehensive systematic literature review as per [16], and developed a scoring scheme for measuring the relevance of software product quality model families for the industrial and scientific communities. In addition, an execution tracing quality study of software product quality models was reported in [3]. Other studies on software product quality models have been carried out by Hegeman in [17] (up to 2011) and by Ferenc et al in [18] (up to 2014), respectively.…”

“…Software product quality models have undergone continuous development for more than 40 years, starting with their conception in the 1970s [1][2][3][4]. Assessing software products involves quality models of a wide range from simple hierarchic decomposition techniques to complex meta-models to cope with the abstract notion of software quality [3][4][5].…”

“…Software product quality models have undergone continuous development for more than 40 years, starting with their conception in the 1970s [1][2][3][4]. Assessing software products involves quality models of a wide range from simple hierarchic decomposition techniques to complex meta-models to cope with the abstract notion of software quality [3][4][5]. Nevertheless, the quality models applied and their concrete implementations differ significantly in terms of the ability to capture the manifestations of software quality.…”