Towards a Fast Cost Estimation Supported by Large Language Models

Hartenstein, S., Johnson, S.L., Schmietendorf, A.

This elaboration describes the conceptual identification of the capabilities of large
language models for cost estimation. For this purpose, the possibilities of the models for
size estimation according to COSMIC Function Point are explained and examined
prototypically. A comparison of a manual measurement with a trained and untrained model
shows the current state of AI capabilities in a concrete example. Finally, the next steps
within the study are briefly discussed.

Appendix with a Generic Software Model by ChatGPT, a Legacy Measurement, a FUR Mapping by ChatGPT, a CFP Measurement by ChatGPT and a FUR Mapping and CFP Measurement by trained ChatGPT.