Software project baseline prediction

Abstract

This paper serves as an informative primer on the agile methodology, offering a foundational understanding of modern software development principles. It explores prevalent methods for SDEE and critically assesses their real‐world viability. Inspired by Tawosi’s work, this paper replicates Tawosi’s findings while introducing the implementation of a variation of his future work mentioned in the original paper. A key focus is an in‐depth examination of both the Deep‐SE architecture and TFIDF/SVM methodology, providing readers with a thorough understanding of their underlying mechanics, capabilities and related work. The paper conducts rigorous experiments, training multiple Deep‐SE and TFIDF/SVM models on a consistent dataset with various permutations. These permutations involve retaining or removing code and hyperlinks from user story descriptions. Results are presented in tables, and implications are discussed in the evaluation chapter, offering insights into how these modifications affect model performance. This informs practical considerations for SDEE and paves the way for future refinements in estimation methodologies.