An UML-based approach to software development cost estimation

CONTEXT: Resource and time estimation is a critical activity in software projects. In literature, various techniques have been proposed, but it's often difficult to perform them, because of data scarcity in the early stages of a project. We studied this issue in the context of a business unit that develops automotive infotainment systems.

GOAL: The purpose of this paper is finding a system to estimate effort and duration for complex embedded software projects, using UML design and a hierarchical "divide and conquer" paradigm.

METHOD: At first, we analyzed and compared estimation methodologies that are available in literature. We found in COCOMO II the model which fitted more our circumstances. By comparing the theoretical estimate model with the actual production process, we produced an UML profile, still based on COCOMO II rationale, but extended and modified, in order to adhere better to our case. The resulting model comes out from a flexible approach, which allows to evaluate effort more thoroughly as we have more detailed information on the estimated project. After model design, we implemented a tool to extract data from the selected UML designer software, and transfer them to a precompiled Excel workbook, conceived specifically for the calculation of project costs using COCOMO II formulae.

RESULTS: The solution has been validated against a completed project. The results are fair in terms of total effort estimate, whereas duration estimate diverges significantly from the actual value. This suggests that the model of effort distribution over time needs to be improved.

CONCLUSIONS: The results suggest that the proposed approach could be a valid option for algorithmic estimation in contexts like the one indicated, at least for total effort assessment. However, the presented solution is far from mature. In the future works, several enhancements could be added to both the model and the tool.