Abstract

This study explores the use of unsupervised machine learning techniques to identify performance patterns and maturity levels among Agile teams. Three clustering algorithms K-Means, DBSCAN, and Hierarchical Clustering were applied to analyze key Agile performance indicators, including sprint efficiency, target achievement, and resource utilization. The models were evaluated using Silhouette Scores to determine clustering quality and interpretability. Results revealed three distinct categories of Agile teams: high-performing, stable, and developing. Among the models, DBSCAN achieved the highest Silhouette Score of 0.728, indicating its superior ability to detect complex, non-linear relationships and identify outlier teams with unique performance behaviors. K-Means and Hierarchical Clustering produced stable and interpretable structures, reinforcing the consistency of the three-cluster configuration. These findings align with Agile maturity theory, suggesting that team performance evolves through identifiable stages of development and process optimization. The study demonstrates that AI-based clustering provides a robust analytical framework for monitoring Agile performance, benchmarking team capabilities, and supporting data-driven managerial decision-making. Integrating machine learning into Agile management practices enhances transparency, fosters continuous improvement, and strengthens organizational adaptability in dynamic project environments.

Keywords: Agile Performance, Machine Learning, Unsupervised Clustering, Agile Maturity, Data-Driven Management

How to Cite:

Durachman, Y. & Rahman, A. W., (2026) “Data-Driven Clustering of Agile Teams Using Unsupervised Machine Learning for Performance Optimization”, Agile Management 1(4), 286-302. doi: https://doi.org/10.63913/ail.v1i4.142