WIT Press

Risk And Software Metrics Datasets


Free (open access)

Paper DOI





764 kb


D.G. Edgar-Nevill


Risk and software metrics datasets D.G. Edgar-NeviH Department of Computer Studies, Napier University, Edinburgh, UK ABSTRACT Many software project assessment and prediction systems are based on the results of analysis gathered from past projects. It is intuitively sensible to gather such data and look for trends upon which to form formulae using statistical techniques. Widely used software metrics systems such as COCOMO [1] and SLIM [2] have been based on results analysed in this way. When building such datasets credibility is usually given to large sets of data rather than smaller. Little regard seems to be given, however, when it is appropriate to add a new projects results to a dataset. Even less often is thought given to when a projects results should be removed from a dataset. This paper considers the problem by analysing the construction and use of historical software project data repositories in a number of case study companies. Guidelines are given on the formation of