This paper proposes an approach for detecting the so- called bad smells in software known as Code Smell. In considering software bad smells, object-oriented software metrics were used to detect the source code whereby Eclipse Plugins were developed for detecting in which location of Java source code the bad smell appeared so that software refactoring could then take place. The detected source code was classified into 7 types: Large Class, Long Method, Parallel Inheritance Hierarchy, Long Parameter List, Lazy Class, Switch Statement, and Data Class. This work conducted analysis by using 323 java classes to ascertain the relationship between the code smell and structural defects of software by using the data mining techniques of Naive Bayes and Association Rules. The result of the Naive Bayes test showed that the Lazy Class caused structural defects in DLS, DE, and Se. Also, Data Class caused structural defects in UwF, DE, and Se, while Long Method, Large Class, Data Class, and Switch Statement caused structural defects in UwF and Se. Finally, Parallel Inheritance Hierarchy caused structural defects in Se. However, Long Parameter List caused no structural defects whatsoever. The results of the Association Rules test found that the Lazy Class code smell caused structural defects in DLS and DE, which corresponded to the results of the Naive Bayes test.
展开▼
