Home / Current Issue / Paper 1709126
Leveraging Transformer-Based Large Language Models for Parametric Estimation of Cost and Schedule in Agile Software Development Projects
Subject area: Science,Engineering and Technology · Area of research: Agile Software Development
Abstract
Accurate estimation of cost and schedule remains a critical challenge in agile software development due to iterative delivery cycles, evolving requirements, and cross-functional team dynamics. Recent advancements in transformer-based large language models (LLMs), such as BERT and GPT, present promising opportunities for improving parametric estimation accuracy through contextual learning and natural language understanding. This paper explores the integration of LLMs into agile project management frameworks to automate and enhance estimation processes based on historical project data, user stories, and sprint planning artifacts. By leveraging pre-trained models fine-tuned on domain-specific repositories, the approach enables predictive modeling of project parameters with improved consistency and scalability. A review of the literature reveals that while traditional machine learning techniques have been used for estimation tasks, LLMs offer superior performance in capturing semantic complexity and stakeholder language. The study further presents a conceptual framework for embedding transformer-based models into agile workflows, highlighting their potential to reduce estimation bias, improve planning accuracy, and facilitate continuous forecasting. This research contributes to the growing intersection between AI-driven software engineering and agile project management, advocating for data-centric decision-making in software delivery environments.
Keywords
Transformer-Based Models, Parametric Estimation, Agile Software Development, Cost and Schedule Forecasting, Large Language Models (LLMs).
How to cite this paper
@article{1709126,
author = {Bamidele Samuel Adelusi, Abel Chukwuemeke Uzoka, Yewande Goodness Hassan, Favour Uche Ojika},
title = {Leveraging Transformer-Based Large Language Models for Parametric Estimation of Cost and Schedule in Agile Software Development Projects},
journal = {Iconic Research And Engineering Journals},
year = {2020},
volume = {4},
number = {4},
pages = {267-278},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1709126.pdf},
abstract = {Accurate estimation of cost and schedule remains a critical challenge in agile software development due to iterative delivery cycles, evolving requirements, and cross-functional team dynamics. Recent advancements in transformer-based large language models (LLMs), such as BERT and GPT, present promising opportunities for improving parametric estimation accuracy through contextual learning and natural language understanding. This paper explores the integration of LLMs into agile project management frameworks to automate and enhance estimation processes based on historical project data, user stories, and sprint planning artifacts. By leveraging pre-trained models fine-tuned on domain-specific repositories, the approach enables predictive modeling of project parameters with improved consistency and scalability. A review of the literature reveals that while traditional machine learning techniques have been used for estimation tasks, LLMs offer superior performance in capturing semantic complexity and stakeholder language. The study further presents a conceptual framework for embedding transformer-based models into agile workflows, highlighting their potential to reduce estimation bias, improve planning accuracy, and facilitate continuous forecasting. This research contributes to the growing intersection between AI-driven software engineering and agile project management, advocating for data-centric decision-making in software delivery environments.},
keywords = {Transformer-Based Models, Parametric Estimation, Agile Software Development, Cost and Schedule Forecasting, Large Language Models (LLMs).},
month = {October},
}