The nested joint clustering via Dirichlet process mixture model.

Area of Science:

• Statistics
• Machine Learning
• Bioinformatics

Background:

• Clustering methods often struggle to model both time-invariant and temporal patterns simultaneously.
• Existing approaches may not effectively capture complex shared patterns among subjects and variables.

Purpose of the Study:

• To propose a novel semi-parametric clustering model using Dirichlet process (DP) mixtures.
• To simultaneously model common and unique temporal patterns within subjects and variables.
• To jointly cluster subjects and variables for enhanced pattern discovery.

Main Methods:

• Utilizing Dirichlet process (DP) mixtures for flexible, non-parametric clustering.
• Developing a semi-parametric model that accounts for both shared and individual patterns.
• Jointly clustering subjects and variables to uncover intricate relationships.

Main Results:

• The proposed DP mixture model demonstrates effectiveness in simulation studies.
• The method successfully identifies novel temporal patterns in wheal size data.
• Simultaneous clustering of subjects and variables reveals complex shared patterns.

Conclusions:

• The novel semi-parametric DP mixture model offers a flexible approach to clustering with temporal data.
• This method enhances the identification of complex patterns in biological and other time-series datasets.
• The joint clustering of subjects and variables provides deeper insights into data structure.