Comparative Analysis of Determination of Method Location between Classes

클래스 간 메소드 위치 결정 방법의 비교

  • 정영애 (단국대학교 전자계산학과) ;
  • 박용범 (단국대학교 전자계산학과)
  • Published : 2006.12.28

Abstract

In Object-Oriented Paradigm, various cohesion measurements have been studied taking into account reference relation among components - like attributes and methods - that belong to a class. In addition, a number of methods have taken into research utilizing manual analysis, that is performed by developer's intuition and experience, and automatic analysis in refactoring field. The verification of objective criteria is demanded in order to process automatic refactoring. In this paper, we propose a method exploiting logistic regression and neural network for analysis of the relationship between six factors considering reference relation and method location among classes. Experimental results demonstrate that the logistic regression predicts the results up to 97% and the neural network predicts the outcomes up to 90%. Hence, we conclude that the logistic regression based method is more effective to predict the method location. Moreover, more than 90% of experimental results from both methods show that the six factors used in Move Method in refactoring are suitable to be used as an objective criteria.

Keywords

Logistic Regression;Refactoring;Determination of Method Location;Neural Network;Cohesion