Evaluating Dependency based Package-level Metrics for Multi-objective Maintenance Tasks

Abstract

Abstract-Role of packages in organization and maintenance of software systems has acquired vital importance in recent research of software quality. With an advancement in modularization approaches of object oriented software, packages are widely considered as re-usable and maintainable entities of objectoriented software architectures, specially to avoid complicated dependencies and insure software design of well identified services. In this context, recently research study of H. Abdeen on automatic optimization of package dependencies provide composite frame of metrics for package quality and overall source code modularization. There is an opportunity to conduct comprehensive empirical analysis over proposed metrics for assessing their usefulness and application for fault-prediction, design flaw detection, identification of source code anomalies and architectural erosion. In this paper, we examine impact of these dependency optimization based metrics in wide spectrum of software quality for single package and entire software modularization. Our experimental work is conducted over open source software systems through statistical methodology based on cross validation fault-prediction and correlation. We conclude with empirical evidence that dependency based package modularization metrics provide more accurate view for predicting fault-prone packages and improvement of overall software structure. Thus, application of these metrics can help the developers and software practitioners to insure proactive management of the source code dependencies and avoid design flaws during software development.

An Empirical Analysis of Approach-Based Metrics Model for Architectural Erosion Detection

“…Focusing on hypothesis 6 (H6), introduced satisfied results in terms of identifying an occurrence of the AE (B = 0.379, t = 6.263, p < 001, F2 = 0.238). Therefore, the findings of the correlation between AM and the emergence of AE are consistent with the observations by 115 , besides, modularization metrics provide a better representation picture for fault prediction, design flaw detection, identifying source code anomalies and architectural degradation 116 , as well as improving architecture 117 with regard to the analysis of faults and changes that could be isolated and separated. Moreover, the results regarding the relationship between ABS and the occurrence of AE (B = 0.215, t = 3.356, p 0.001, F2 = 0.075) are consistent with the analysis and examination of the various study on architectural smells and their relationship to determine the presence of architectural erosion 55,56,59,60,118 and instability in order to identify hidden defects in software architecture 119 .…”

“…Inconsistencies between intended architecture as detected as symptoms of architectural degradation are investigated by many source code metrics such as lack of method cohesion metric 15 and lack of cohesion metric for packages 31,38 . Dependency optimization-based metrics are used to identify source code anomalies and architectural erosion through package cohesiveness quality metric to examine internal package dependencies 39 and tight class cohesion 40 . Correlations and interactions between architecture changes and decay 27 are significant concept to measure actual cluster interactions to possible cluster interactions using the ratio of cohesive interactions metric.…”

“…The term "architecture modularization" is used to describe the degree to which a system or piece of software is divided into independent modules that have little effect on one another when modified. In studies 39,41 , proposed package quality measures (inter-package modularization dependencies, inter-package modularization connections, inter-package modularization cyclic dependencies, and inter-package modularization cyclic connections) over a wide range of software quality, including vulnerability detection, fault-proneness, and violation of coding standard. For the purpose of figuring out if the quality of software architecture is sufficient and if it is degrading over time.…”

“…Analysis of the dependency between conceptual architecture on the one hand and architecture changes, erosion, the presence of vulnerabilities, problem violations, and defect-related activity on the other was one of most significant motivations and areas of interest for researchers in many studies. 15,27,31,35,39,45,61,[63][64][65][66] , which adopted coupling metrics between models, objects, packages, and classes to determine the state of the system architecture in terms of evolution or degradation. For example, coupling metrics such as coupling between objects, afferent coupling, efferent coupling, data abstraction coupling, number of coupled classes, sum of coupling, module dependency strength (MDS), and dependency frequency.…”