Dynamic Reusability Measurement Using Machine Learning Algorithms in Object-Oriented Environment

Abstract

Software reusability is measured with the help of object-oriented static and dynamic metrics to better predict the quality of software effectively. Most of the research exists in the literature to measure the reusability of software based on static metrics by capturing the design patterns that do not take a count on run time behaviour of object-oriented software. Therefore, in this paper, authors derived a new approach to measure the software reusability of a design pattern based on dynamic metrics that capture the run time features of object-oriented programming, i.e., run time polymorphism, dynamic binding, etc. Besides, dynamic metrics were extracted by using AspectJ, an implementation of aspect-oriented programming on the eclipse platform. Further, authors proposed a fuzzy logic model based on static and dynamic metrics to measure reusability factor of a design pattern and compared it with various other existing machine learning algorithms such as linear regression (LR), Gaussian processes (GP), simple linear regression (SLR), SMOreg, random forest (RF) and M5P. In the end, an experimental study was done on 18 design patterns, and it is concluded that proposed fuzzy logic approach gives satisfactory results based on dynamic metrics to better predict the reusability factor of a design pattern.KeywordsFuzzyReusabilityDynamicStaticMachine learning