Abstract:
Computer software has become an important foundation in several versatile domains including medicine, engineering, etc. Consequently, with such widespread application of software, the essential need of ensuring certain software quality characteristics such as efficiency, reliability and stability has emerged. In order to measure such software quality characteristics, we must wait until the software is implemented, tested and put to use for a certain amount of time. Several software metrics have been proposed in the literature to avoid this long and costly process, and they proved to be a good means of estimating software quality. For this purpose, software quality prediction models are built. These are used to establish a relationship between internal sub-characteristics such as inheritance, coupling, size, etc. and external software quality attributes such as maintainability, stability, etc. Using such relationships, one can build a model in order to estimate the quality of new software systems. Such models are mainly constructed by either statistical techniques such as regression, or machine learning techniques such as C4.5 and neural networks. We build our model using machine learning techniques in particular rule-based models. These have a white-box nature which gives the classification as well as the reason for it making them attractive to experts in the domain.
In this thesis, we propose a novel heuristic based on Artificial Bee Colony (ABC) to optimize rule-based software quality prediction models. We validate our technique on data describing maintainability and reliability of classes in an Object-Oriented system. We compare our models to others constructed using other well established techniques such as C4.5, Genetic Algorithms, Simulated Annealing, Tabu Search, multi-layer perceptron with back-propagation, multi-layer perceptron hybridized with ABC and the majority classifier. Results show that, in most cases, our proposed technique out-performs the others in different aspects.
