PCI-PSO: Preference-Based Component Identification Using Particle Swarm Optimization

Abstract

Abstract Nowadays, component identification is one of the main challenges of software analysis and design. The component identification process aims at clustering classes into components and subcomponents. There are a number of methods to identify components in the literature; however, most of them cannot be customized to software architect’s preferences. To address this limitation, in this paper, we propose a preference-based method by the name of preference-based component identification using particle swarm optimization (PCI-PSO) to identify logical components. PCI-PSO provides a novel method to handle the software architect’s preferences using an interactive (i.e. human in the loop) search. PCI-PSO employs a customized PSO to automatically classify classes into suitable logical components and avoid the problem of identifying the proper number of components. We evaluated the effectiveness of PCI-PSO with four real-world cases. Results revealed that PCI-PSO has an ability to identify more cohesive and independent components with respect to the software architect’s preferences in comparison to the existing component identification methods.