| المستخلص: |
Software Quality is an important issue in the development and success of the software. It is concerned with modifications and improvements necessary to meet the evolving needs and performed during maintenance phase of Software Development Life Cycle (SDLC). The problem that is accompanied to any modification is the possible low quality of the resulted software. Large class bad smells are serious design flaws that could affect the code’s quality attributes such as understand ability and readability. These flaws could ultimately lead to difficulties in maintaining the code and adding new functionalities. This work aims to detect large class bad smells automatically to help developers and engineers to detect large class bad smells from the get-go. This support keeping the code clean and easy to be understood, thus eliminating the need to constantly referring back to the documentation every time we try to add or repair functionality. Usually, the large class bad smell is identified by using the coupling and cohesion metrics and compared to the identified class smelly elements to determine if one or more large class bad smells exist. Large Class Smell Detection (LCSD), is a proposed approach used in this work to automate the development of a large class bad smell detection model that is based on cohesion and coupling metrics. The automation of this development utilizes Genetic Algorithm (GA) and Artificial Neural Network (ANN). LCSD’s results showed that its performance is very good in finding large class bad smells. The correctness of LCSD has been measured by using binomial technique, and achieved high results, which is 96.67%.
|
