Goal-Centric Traceability for Managing Non-Functional Requirements

Author(s): Cleland-Huang, J., Settimi, R., BenKhadra, O.
Venue: ICSE '05
Date: 2005

Type of Experiement: Controlled Experiment

Quality
3

Link: http://portal.acm.org/citation.cfm?id=1062455.1062525

This paper describes a method for automated tracing of UML classes to non-functional requirements. The authors describe a technique called Goal-Centric Traceability (GCT), which uses traceability information to perform change impact analysis on non-functional requirements (NFRs). NFRs are modeled in a softgoal interdependency graph (SIG), which captures conflicts between NFRs and tradeoffs of design decisions. By tracing classes to nodes in the SIG, change impact analysis can easily determine which NFRs may be impacted by a change to the design or code.

GCT utilizes a probabilistic information retrieval method to generate links between UML and SIG elements. This paper studies the effectiveness of GCTs method by applying it to the Ice Breaker System. The Ice Breaker System comprises 180 SIG elements (82 goals and subgoals and 98 operationalizations) and 75 classes. The authors manually constructed a traceability matrix against which they can measure the recall and precision of GCT. GCT's IR method was trained on a subset of the data (25 classes and 60 SIG elements) to determine the optimal threshold value, which was 0.02. Then, GCT was run on the entire data set.

On the training set, GCT obtained recall and precision of 95% and 42%, respectively. On the entire data set, GCT obtained recall and precision of 87% and 51%, respectively. Compared to experiments run on other artifacts from the same dataset, GCT returned higher precision for equivalent levels of recall. The authors note that when constructing the traceability matrix, they considered any "gray" classes as true links. Their tool tends to do the same, so this decision improves their precision measures.

0