Detecting Bad Smells with Machine Learning Algorithms: an Empirical Study

Complementary Webpage

Dataset Download

Class Level

• God Class
• Refused Parent Bequest

Method Level

• Long Method
• Feature Envy

List of Systems

• apache-commons-lang
• apache-commons-codec
• apache-commons-io
• apache-commons-logging
• apache-lucene
• checkstyle
• hadoop
• hibernate
• htmlunit
• jasperreports
• jfreechart
• jhotdraw
• jmeter
• quartz
• spring-framework
• squirrelsql
• struts
• tapestry
• tomcat
• weka

Manual Validation Questions

The answer after the question defines which answer indicates the presence of Bad Smell.

God Class

• Does the class have more than one responsibility? Yes
• Does the class have functionality that would fit better into other classes? Yes
• Would splitting up the class improve the overall design? Yes

Reference: Building Empirical Support for Automated Code Smell Detection. Jan Schumacher et al.

Refused Parent Bequest

• Does the class use only a little of parent’s behaviour? Yes
• Does the parent class provide more than a few protected members? Yes
• Does the class is too small/simple? No

Reference: Object-Oriented Metrics in practice. Michele Lanza, Radu Marinescu. (Book)

Long Method

• Does the method have many conditional branches? Yes
• Does the method is excessively large? Yes
• Does the method use many variables? Yes

Reference: Object-Oriented Metrics in practice. Michele Lanza, Radu Marinescu. (Book)

Feature Envy

• Does the method use directly more than a few attributes of other classes? Yes
• Does the method use far more attributes from other classes than its own? Yes
• Do the used “foreign” attributes belong to very few other classes? Yes

Reference: Object-Oriented Metrics in practice. Michele Lanza, Radu Marinescu. (Book)