FULSOME: fuzzy logic for software metric practitioners and researchers

Abstract

There has been increasing interest in recent times for using fuzzy logic techniques to represent software metric models, especially those predicting the software development effort. The use of fuzzy logic for this application area offers several advantages when compared to other commonly-used techniques. These include the use of a single model with different levels of precision for the inputs and outputs used throughout the development life-cycle, the possibility of model development with little or no data, and its effectiveness when used as a communication tool. The use of fuzzy logic in any applied field, however, requires that suitable tools are available for both practitioners and researchers-satisfying both interface- and functionality-related requirements. After outlining some of the specific needs of the software metrics community, including results from a survey of software developers on this topic, this paper describes the use of a set of tools called FULSOME (FUzzy Logic for SOftware MEtrics). The development of a simple fuzzy logic system by a software metrician and its subsequent tuning are then discussed using a real-world set of software metric data. The automatically generated fuzzy model performs acceptably when compared to regression-based models.