Evaluating Aspect Mining Techniques: A Case Study

Abstract

Aspect mining aims at identifying cross-cutting concerns in existing systems and therefore advocates the adaption to an aspect-oriented design. This paper presents a case study examining three existing aspect mining techniques from the literature by applying them to four different open source java applications. We compare and evaluate the individual technique and confirm the findings of a previous study of combining different aspect mining techniques in order to get better results with less manual intervention. 15th IEEE International Conference on Program Comprehension (ICPC'07) 0-7695-2860-0/07 $20.00 © 2007 15th IEEE International Conference on Program Comprehension (ICPC'07) 0-7695-2860-0/07 $20.00

Automated Aspect Recommendation through Clustering-Based Fan-in Analysis

“…We evaluated it on two completely different systems, C-based Linux, and a Java system, JHotDraw 2 , which has been evaluated in many previous aspect mining research work [4], [10], [12]. Through detailed evaluation of several CCCs in Linux and JHotDraw, we demonstrate that CBFA can provide accurate recommendations while improving aspect mining coverage significantly compared to other state-of-the-art mining approaches.…”

“…Since it is designed as an exercise to show how to use design patterns, it was analyzed by many mining research groups to evaluate their mining approaches [4], [10], [12]. The version we analyzed is 5.4b, which includes about 12K lines of code.…”

“…We choose five widely analyzed CCCs [4], [10], [12] in JHotDraw, which include: 1) Undo: an undo concern in a graphical editor like JHotDraw performs undo and redo activities. It is an implementation of the Command design pattern [17].…”

“…As reported in [10], [12], choosing a suitable threshold has a significant impact on the mining performance of fan-in analysis.…”

Evaluation Measures for Partitioning based Aspect Mining Techniques

“…There are very few comparisons made between the aspect mining techniques proposed so far [4,10,16,17], and even less comparisons based on the obtained results [4,17]. One important cause is the lack of measures for evaluating the results obtained and the quality of the results (i.e.…”