An international research group representing leading scientific centers in Azerbaijan and Australia has proposed a novel approach to a long-standing fundamental problem in artificial intelligence and data mining.
AZERTAC reports that the article titled "Absolute Indices for Determining Compactness, Separability, and Number of Clusters," presenting the research outcomes, has been published in Pattern Recognition, one of the world's most influential scientific journals in artificial intelligence.
Clustering, grouping large datasets based on similar features, plays an important role in artificial intelligence. However, existing indices used to evaluate clustering quality (compactness and separability) are relative in nature, primarily used to compare different algorithms, and often yield contradictory results for complex data structures. In this study, the researchers created a framework for absolute evaluation that overcomes these limitations. The paper introduces two novel indices for evaluating clustering outcomes: one measuring cluster compactness and the other assessing cluster separability.
Grounded in rigorous mathematical foundations, this innovation extends beyond theoretical value, demonstrating high potential for practical applications across various fields. For instance, it can make significant contributions to medicine through precise disease classification and grouping of patients by drug sensitivity; cybersecurity by enabling real-time detection of cyberattacks and network anomalies; smart cities by optimizing traffic and pedestrian flow; space research by facilitating high-precision analysis of satellite imagery and geophysical maps; as well as agriculture and beyond.
The international team behind this world-class scientific achievement includes Professor Adil Bagirov and Associate Professor Nargiz Sultanova (Federation University Australia), Corresponding Member of ANAS Ramiz Aliguliyev (Head of Department at the Institute of Information Technology, Director of the Data Science and Artificial Intelligence Institute at Azerbaijan Technical University), and Associate Professor Sona Taheri (Royal Melbourne Institute of Technology).
© All rights reserved. When using these materials, reference to www.iсt.az is required.