Automating the Estimation of Project Size from Software Design Tools Using Modified Function Points

Final year students in the Bachelor of Computing complete an industry project where they work in teams to build an IT system for an external client. Grading projects in these circumstances is difficult because of the huge variability of projects and clients. A method of ameliorating some of the variation is to perform a function point count on the projects. Due to the large number of projects and the changing scope of projects a method of automatically counting function points has been devised that uses the output from design tools that students have used. Principally the method counts use cases and database tables. The method has been successful in that no statistical difference in function point counts was found regardless of the implementation environments of systems. However, the first function point count produced during the design phase resulted in values that are lower than expected. The reason for this is that there are omissions from the design. The students will perform another at the user testing stage. The average function point count is 270 with a standard deviation of 130. Currently, the method also assumes that the students are following a traditional waterfall development model. The paper discusses two issues (a) proposing a metric for project size and (b) automating the production of that metric.