Estimating software size with UML models

  • Authors:

  • Ghislain Levesque;Valery Bevo;De Tran Cao

  • Affiliations:

  • University of Quebec at Montreal, Montreal;University of Quebec at Montreal, Montreal;University of Cantho, Cantho, Vietnam

  • Venue:
  • Proceedings of the 2008 C3S2E conference
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Function points have been used for more than 25 years for estimating software size and building productivity models. Today, three methods to do it are accepted as ISO standards. The theory behind this type of measurement is more explicit but none of these methods have yet been fully automated. All of them still require the involvement of an expert in order to be used correctly. Why is it so difficult to implement those methods? In our opinion, the main reason lies in the fact that each method has its own vocabulary and its own way of modeling software. The research work presented here has been realised mainly through two doctoral theses, one trying to automate the measure from a UML perspective and the other to add an objective measure of complexity to a standard measure in COSMIC-FFP in order to reach a higher level of confidence with those measures. So far, it can be concluded that, from UML use-cases and Actor-Object sequence diagrams of a system application, the number of messages exchanged correspond to the number of function points according to the COSMIC-FFP method. Going farther and adding the number of conditions or decisions to be applied according to UML version 2.0 would add more precision taking into account the complexity of the processes.