Decision tree-based Design Defects Detection

Design defects affect project quality and hinder development maintenance. Consequently, experts need to minimize these in software systems. A promising approach is apply the concepts of refactoring at higher level abstraction based on UML diagrams instead code level. Unfortunately, we find literature many that are described textually there no consensus how decide if a particular design violates model quality. Defects could be quantified as metrics rules represent combination metrics. However, it difficult manually best threshold values for In this paper, propose new identify using ID3 decision tree algorithm. We aim create each defect. experimented our four defects: The Blob, Data class, Lazy class Feature Envy defect, 15 Object-Oriented generated give very detection results open source projects tested paper. Lucene 1.4 project, found precision 67% recall 100%. general, accuracy varies from 49%, reaching 80%.