Empirical Validation of Information Theory-Based Software Metrics

点击翻译按钮获取中文摘要

CCR-0098024Empirical Validation of Information Theory-Based Software MetricsEdward B. Allen, P.I, and Rayford B. Vaughn, co-P.I.Empirical Validation of Information Theory-Based Software Metrics is aresearch project of Mississippi State University that is empiricallyvalidating the usefulness of a new generation of software metrics, basedon information theory.Software engineers employ a wide variety of diagrams during developmentof software. Because many abstractions of software are represented bygraphs, metrics of graph attributes have the potential for wideapplication. Information theory is an alternative to counting, focusingon the amount of information in an attribute. The contribution of thisproject to the state of the art is empirical evidence that informationtheory-based software metrics of size, length, complexity, coupling andcohesion can be useful timely predictors of software quality, and thatthey have advantages over counting-based metrics.Case studies in collaboration with industrial and government softwaredevelopment organizations are providing a meaningful evaluation byexamining a variety of real-world software systems large enough to becomparable to other industry projects. Collaborators include EDS Inc.,Ericsson Inc., and MPI Software Technology Inc. Results will bedisseminated throught the Center for Empirically Based SoftwareEngineering (CeBASE). The empirical evidence generated by this projectis aimed to facilitate a new level of cost-effective improvement tosoftware quality.