Multivariate functional data clustering using adaptive density peak detection.

Area of Science:

• Statistics
• Data Science
• Machine Learning

Background:

• Clustering multivariate functional data is complex due to data residing in infinite-dimensional spaces.
• Existing methods often require iterative processes and can be computationally intensive.

Purpose of the Study:

• To propose a novel, computationally fast clustering method for multivariate functional data.
• To introduce an adaptive density peak detection technique for efficient cluster center identification.

Main Methods:

• Utilized adaptive density peak detection based on functional density estimates and distance to higher-density neighbors.
• Proposed two functional density estimators: k-nearest neighbor on L2 distance or functional principal components, and k-nearest neighbor on functional principal scores.
• Developed a user-friendly R package (FADPclust) for public access.

Main Results:

• The proposed clustering method is computationally efficient, avoiding iterative procedures.
• Simulation studies demonstrated the flexibility and advantages of the method compared to existing approaches.
• The method was successfully applied to a real-world lung cancer research case study.

Conclusions:

• The adaptive density peak clustering method offers a fast and effective solution for multivariate functional data.
• The FADPclust R package facilitates the application of this novel technique in research.
• The method shows promise for applications in complex scientific domains like medical research.