Mesoscopic analysis of networks: applications to exploratory analysis and data clustering

Area of Science:

• Complex systems analysis
• Network science
• Data mining

Background:

• Correlation matrices are crucial for understanding complex systems.
• Analyzing mesoscopic structures in large datasets presents challenges.
• Existing methods may not fully capture hierarchical data organization.

Purpose of the Study:

• To adapt modularity-based algorithms for correlation matrix analysis.
• To explore the multiresolution topological structure of data.
• To demonstrate the algorithms' performance on biological and benchmark datasets.

Main Methods:

• Application of modularity-based algorithms from complex network theory.
• Utilizing multiresolution analysis to identify clusters at various topological levels.
• Testing on neural connectivity data of *Caenorhabditis elegans* and the Iris dataset.

Main Results:

• Successful adaptation of modularity algorithms for correlation matrices.
• Identification of data clusters across different hierarchical levels.
• High performance in analyzing *C. elegans* neural networks and classifying the Iris dataset.

Conclusions:

• Modularity-based algorithms offer a robust framework for mesoscopic analysis of correlation matrices.
• The multiresolution approach reveals hierarchical structures effectively.
• Demonstrated utility in both biological network analysis and machine learning classification tasks.