Predicting Object Oriented Software Systems Maintainability At Design Level Using K Means Clustering Technique

Vol.6 No.4

Year: 2012

Issue: April-June

Title: Predicting Object Oriented Software Systems Maintainability At Design Level Using K Means Clustering Technique   

Author Name: Dr. Anil Kumar Malviya, VINOD KUMAR YADAV   

Synopsis: 

Software maintenance is single most expensive activity in entire software development. One way to control the maintenance cost is to utilize software metrics during design phase of development. This paper examined application of K-means clustering technique for identifying the maintainable classes using object-oriented metrics. In this work data clustering technique’s K-means clustering is used to evaluate a software system’s maintainability of Object oriented system based model mainly UIMS (User Interface Management System and QUES (Quality Evaluation System) class’s data. Among the clustering techniques, K-means or Partition clustering will construct non overlapping groups. In this paper we are not only present preliminary experimental work of software maintenance using software metrics for the sample data is being simulated on Matlab but also present the significant level of goodness of clusters using Chi-square Test. . Experimental results on MatLab shows that the algorithm is able to decide the cluster with goodness of fit among clusters using Chi-Square Test that provides the help to the software designers and maintainers to take the appropriate action at design level. It can also be used by software designer to change or modify the design of difficult to maintain classes at design level of software.