A Deep Hybrid Learning Based Story Point Estimation for Agile Project

Abstract

The Agile technique is a widely recognized and used strategy within the field of software engineering. Numerous organizations that prioritize adaptability, customer-centric methodologies, and team appreciation use agile practices. The use of agile approach in software development is undeniably associated with the capacity to adapt to changing environments and foster personal growth among employees. The Agile methodology allows for the active participation of all team members, fostering creativity and facilitating the management of large-scale projects within expansive teams. The proficiency in inter-team and cross-team communication for the purpose of deliberating project resolutions and effectively implementing effective solutions to real-world challenges. In the realm of agile methodology, a story point refers to a challenging job that has the potential for adverse consequences on team collaboration and overall project performance, often resulting in budgetary overruns. This relatively small but crucial area need meticulous deliberation throughout the estimating process, when a project manager with substantial expertise is tasked with making pivotal decisions for the project. If the manager is deficient in certain expertise, it might have catastrophic consequences for the whole business. Story points refer to the user requirements that are carefully considered by a business analyst, with potential involvement from a software developer to clarify the expectations. Projects that have clear and well-defined timetables are more likely to be completed on time, resulting in a higher level of satisfaction in terms of team performance. Artificial Intelligence (AI) has emerged as a rapidly expanding field in the current decade, with several activities being performed utilizing this advanced technology. Deep Learning, Federated Learning, ML, Reinforcement learning, and several other techniques are integral components of artificial intelligence (AI). The practice of artificial intelligence (AI) in addressing practical challenges partakes provided researchers with an opportunity to apply their repertoire of knowledge and skills towards resolving environmental issues. This is a significant demonstration of software engineering, whereby automation is used to enhance human productivity and save their precious time. This phenomenon is seen in the context of narrative point estimate with artificial intelligence (AI), whereby AI is used to provide precise estimations in order to enhance project planning. This study introduces a unique Deep Hybrid Learning Model, namely GPT2-CNN, which is used as an Agile Project Estimation approach through story point. The Deep Hybrid Learning Model, GPT2-CNN, utilizes a GPT2 etymological model that has prior experience, combined with deep neural-based architecture known as CNN. This integration enables our models to effectively comprehend the v interconnections among words, taking into account the contextual information surrounding a specific word and its placement within the sequence. The findings of our study indicate that our proposed Hybrid Learning Model, namely GPT2-CNN, has a median Mean Absolute Error (MAE) of 1.96. This performance surpasses that of the current baseline technique used for estimating within-project estimates. The ablation research also demonstrates the efficacy of our deep hybrid learning architecture in augmenting the estimation process of agile user stories, henceforth rank the momentous progressions of AI in Agile story point estimation. Our findings are substantiated by the implementation of five distinct hybrid models, namely GPT2-CNN (with default settings), GPT2-LSTM, GPT2-LSTM-CNN (a fusion of three deep learning models), and RoBERTa-RoBERTa. The whole of the experimental configuration confirms the strong performance of GPT2-CNN(AdamW). The task of estimating story points is regarded as a complex endeavor, as noted by Agile practitioners. The use of big data and powerful computers is expected to significantly improve the quality of the solution offered.