Visualization and Analysis in Automated Trace Retrieval

Abstract

This paper describes a new visualization technique for helping analysts to understand the potential impact of changing requirements, and for providing valuable early feedback on the quality of a software design. Based on the candidate links that are automatically generated by a trace retrieval tool, VisMatrix creates a graphical representation of the requirements trace matrix showing not only where candidate links exist, but also the strength of those links. New metrics derived from the trace visualization, and its underlying trace matrix, are introduced. These include 'trace clustering' for evaluating modularity, and 'trace scope' for analyzing fan-in and fan-out behavior of traces, where fan-in represents the number of requirements influencing a design artifact, and fan-out represents the degree of influence a single requirement has on the set of design artifacts. Examples are drawn from the automatically generated trace matrices of four different data sets to illustrate the visualization techniques and analyze the proposed metrics.