A comparison of functionality-based packaging using ga and adaptive knn clustering as two approaches to package software

In large scale object-oriented (OO) software, package forms the essential structural component of the system. Packaging operation aims at grouping OO classes to provide well-identified functions to the rest of the software. It is meant to improve the software quality such as high maintainability and high reusability. Most of the packaging approaches are based on an optimal balance between cohesion and coupling quality attributes. In this article, we compare of two packaging approaches: functionality-based packaging relied on genetic algorithm, GA (measured with a metric named OverallPackaging) and adaptive k-nearest neighbor (A-KNN) clustering. The comparison was done in terms of cohesion and coupling at package level. Both approaches are applied on an open-source system called Trama. In term of cohesion and coupling at the package level, both approaches achieved good results compared with the original packaging of Trama. However, the A-KNN-based approach achieved better results than OverallPackaging.