Accurate software effort estimation is a critical and challenging activity in Agile Software Development, as it directly impacts project cost, schedule, resource allocation, and overall success. Traditional estimation techniques such as Expert Judgment, Planning Poker, COCOMO, Function Point Analysis, and Use Case Points often struggle to deliver reliable estimates in dynamic and evolving Agile environments. With the growing availability of historical project data, supervised Machine Learning (ML) techniques have emerged as effective data driven alternatives for improving estimation accuracy and consistency. This paper presents a Systematic Literature Review (SLR) of recent supervised ML approaches applied to Story Point–based software effort estimation in Agile projects. The review analyzes regression models, ensemble learning techniques, neural networks, and emerging Natural Language Processing (NLP) and Transformer based models, including Random Forest, Gradient Boosting, Support Vector Machines, GPT2SP, and Het- eroSP. The findings indicate that ensemble learning approaches consistently outperform conventional estimation methods in terms of robustness and predictive accuracy, while Transformer based models show strong potential for automated Story Point estimation using textual user story information. The study also identifies key research challenges related to dataset scarcity, estimation subjectivity, model explainability, and cross project generalization. Finally, future research directions are discussed, emphasizing Explainable AI, hybrid estimation frameworks, and intelligent Agile analytics..

Software effort estimation is a critical activity for Project Managers, Developers, Testers, and Business Analysts because it supports the prediction of software effort, schedule, budget, and required resources for successful project delivery. Inaccu- rate estimation frequently results in project failure, schedule overruns, and budget imbalance. Traditional software estimation models use sizing metrics such as KLOC, Function Point Analysis (FPA), Use Case Points (UCP), and Test Case Points. However, these ap- proaches often fail to provide accurate estimation results in Agile environments. Consequently, modern Agile methodolo- gies increasingly rely on Story Points derived from User Stories as sizing metrics. This Systematic Literature Review demonstrates that super- vised Machine Learning techniques are widely applied for Story Point–based software effort estimation. The reviewed studies investigated various models including Linear Regres- sion, Support Vector Regression (SVR), Decision Tree Regres- sion, Random Forest, Gradient Boosting, Deep Learning, and Transformer-based approaches. The studies evaluated several performance metrics including MAE, MMRE, MdMRE, RMSE, and PRED(n). Experimental evidence suggests that ensemble methods such as Random Forest and Gradient Boosting generally provide superior esti- mation accuracy compared with conventional approaches. Future research should focus on developing hybrid esti- mation frameworks that combine Expert Judgment with Ma- chine Learning techniques. In addition, integration with Agile project management platforms such as Jira and GitHub can support continuous model retraining and real-time estimation feedback. Future work should also investigate Explainable AI (XAI), Transfer Learning, Federated Learning, and human-in- the-loop estimation models to improve estimation robustness, interpretability, and industrial applicability.

[1]B. W. Boehm, Software Engineering Economics. Springer, 2002. [2]P. Musilek, W. Pedrycz, N. Sun, and G. Succi, “On the sensitivity of COCOMO II software cost estimation model,” in Proc. IEEE Symposium on Software Metrics, 2002. [3]A. Z. Abualkishik and L. Lavazza, “IFPUG function points to COSMIC function points convertibility,” Information and Software Technology, vol. 97, pp. 179–191, 2018. [4]P. Faria and E. Miranda, “Expert judgment in software estimation during the bid phase,” in Proc. IWSM-MENSURA, 2012. [5]J. Grenning, “Planning poker or how to avoid analysis paralysis while release planning,” Renaissance Software Consulting, 2002. [6]S. M. Satapathy and S. K. Rath, “Empirical assessment of machine learning models for agile effort estimation using story points,” Springer, 2017. [7]V. Van Hai et al., “Toward improving the efficiency of software de- velopment effort estimation,” IEEE Access, vol. 10, pp. 83249–83264, 2022. [8]M. Rahman, T. Gonc¸alves, and H. Sarwar, “Review of existing datasets used for software effort estimation,” International Journal of Advanced Computer Science and Applications, vol. 14, no. 7, 2023. [9]M. A. Ramessur and S. D. Nagowah, “A predictive model to estimate effort in a sprint using machine learning techniques,” International Journal of Information Technology, 2021. [10]M. Ferna´ndez Diego et al., “An update on effort estimation in agile software development: A systematic literature review,” IEEE, 2020. [11]H. Phan et al., “Heterogeneous graph neural networks for software effort estimation,” in ACM/IEEE ESEM, 2022. [12]M. Fu and C. Tantithamthavorn, “GPT2SP: A transformer-based agile story point estimation approach,” IEEE, 2023. [13]S. M. Satapathy and S. K. Rath, “Empirical assessment of machine learning models for agile software development effort estimation using story points,” Springer-Verlag London Ltd., 2017. [14]Z. Z. K., S. K. Tipu, and S. K. Zia, “An effort estimation model for agile software development,” Advances in Computer Science Applications, vol. 2, no. 1, pp. 314–324, 2012. [15]M. A. Ramessur and S. D. Nagowah, “A predictive model to estimate effort in a sprint using machine learning techniques,” International Journal of Information Technology, 2021. [16]Q. Bushra and A. Kadam, “An improved technique for software cost estimations in agile software development using soft computing tech- niques,” IT in Industry, vol. 9, no. 2, 2021. [17]A. G. Duszkiewicz et al., “On identifying similar user stories to support agile estimation based on historical data,” in Agile-ISE 2022 Interna- tional Workshop on Agile Methods for Information Systems Engineering, Leuven, Belgium, June 2022. [18]W. Rosa and S. Jardine, “Data-driven agile software cost estimation models for DHS and DoD,” Journal of Systems and Software. [19]E. Rodr´ıguez Sa´nchez, E. F. Va´zquez Santacruz, and H. Cervantes Maceda, “Effort and cost estimation using decision tree techniques and story points in agile software development,” Mathematics, vol. 11, p. 1477, 2023. [20]H. Phan et al., “Heterogeneous graph neural networks for software effort estimation,” in ACM/IEEE International Symposium on Empirical Software Engineering and Measurement (ESEM), Helsinki, Finland, Sept. 2022. [21]H. Phan et al., “Story point level classification by text level graph neural network,” in 2022 IEEE/ACM International Workshop on Natural Language-Based Software Engineering (NLBSE), 2022. [22]M. Fu and C. Tantithamthavorn, “GPT2SP: A transformer-based agile story point estimation approach.”

© 2025 International Journal of Engineering and Techniques (IJET).

Digital Object Identifier (DOI)

IJET assigns a unique DOI to every accepted article via Zenodo for persistent linking and citation. Your article’s DOI: https://doi.org/{{doi}}

Open DOI About Zenodo DOI

Close