Measuring the Quality of Design Pattern Detection Results

Author(s): Shouzheng Yang ; York Univ., Toronto, ON, Canada ; Manzer, A. ; Tzerpos, V.
Venue: International Conference on Software Engineering
Date: 2-6 March 2015

Type of Experiement: Survey/Multi-Case Study
Class/Experience Level: Professional
Data Collection Method: Observation, Code Metric


This paper discusses a new approach to evaluate results of design pattern detection tools. It is often the case that every tool will return a different number of design patterns used. This gets even more problematic when classes are involved. In the current world of object oriented system, some might count a factory which produces objects from an interface type a single instance. This might be problematic, because others might count it multiple instances; one instance for every concrete object type it can produce.

The researches preformed many case studies on their metrics (GAP and GAR). These metric aggregates some data from other pattern detection tools to determine which tools are accurate and correlate all the findings into a design pattern detection system with a higher detection, accuracy and precision rate. Their system also is better at gauging details of a patterns concrete implementation due to DAP and DAR. These metrics aid with detection of all concrete classes such as any concrete Shape (Square, Circle...) created by a ShapeFactory.