COMPARATIVE ANALYSIS OF MODEL PERFORMANCE FOR TIME ESTIMATIONS IN AI PROJECTS

Abstract

Time estimation is an important part of project planning, especially in Artificial Intelligence (AI) projects, which show unique complexities and variability. This study looks into time estimation using machine learning models trained on a novel dataset of AI projects. By comparing the performance of transformer-based models and Multilayer Perceptron (MLP) architectures, this paper highlights the benefits of domain-specific modeling. The results show that the transformer models are superior in handling complex AI datasets. Discussed results are for the basic version of these models and the paper highlights multiple potential improvements to increase the future performance.

References

1. Boehm, B. (1988). "A Spiral Model of Software Development and Enhancement".
2. Albrecht, A. J., & Gaffney, J. E. (1983). "Function Point Analysis for Software Complexity".
3. Tran, D., & Nguyen, Q. (2024). "Framework for AI-Based Models in Project Planning and Effort Estimation Accuracy".
4. Jørgensen, M., & Shepperd, M. (2007). "A Systematic Review of Machine Learning Techniques in Software Project Management".
5. Amershi, S., Begel, A., Bird, C., & Zimmermann, T. (2019). "AI Workflows: Iterative Development and Data Dependency".
6. Mockus, A., Fielding, R. T., & Herbsleb, J. D. (2003). "Challenges in Dynamic and Exploratory Software Development".