Adaptive tree-reinforced clustering using hierarchical social relations and Q-learning in dynamic data environments

Area of Science:

• Data Science
• Artificial Intelligence
• Network Engineering

Background:

• Clustering large-scale, heterogeneous, and evolving data is a significant challenge.
• Traditional clustering methods struggle with computational complexity, noise, and adaptability.

Purpose of the Study:

• Introduce the Adaptive Tree-Reinforced Clustering (ATRC) framework.
• Address limitations of traditional clustering models in complex data systems.

Main Methods:

• Hybrid model fusing Tree Social Relation (TSR) for hierarchical clustering and Q-Learning for reinforcement optimization.
• TSR establishes initial clusters; Q-Learning refines assignments to avoid local minima.

Main Results:

• ATRC demonstrated superior performance on a heterogeneous IoT sensor network.
• Achieved up to 13% higher packet delivery rates, 28% lower energy consumption, and 0.9x faster convergence compared to benchmarks.
• Enhanced stability and accuracy in uncertain/noisy conditions.

Conclusions:

• Integrating reinforcement learning with hierarchical modeling offers a scalable and energy-efficient clustering solution.
• ATRC is suitable for real-time IoT, fog computing, and data-intensive intelligent systems.