Semisupervised Clustering by Iterative Partition and Regression with Neuroscience Applications

Area of Science:

• Statistical learning
• Data mining
• Artificial intelligence
• Neuroscience

Background:

• Regression clustering integrates unsupervised and supervised learning.
• It has broad applications in AI and neuroscience.
• Practical use requires determining cluster number, data labels, and regression coefficients.

Purpose of the Study:

• Review estimation and selection issues in regression clustering.
• Introduce a model selection technique for determining the number of clusters.
• Develop a computational procedure for regression clustering estimation and selection.

Main Methods:

• Review of least squares and robust statistical methods for regression clustering.
• Development of a model selection technique for cluster number determination.
• Creation of a computing procedure for estimation and selection.

Main Results:

• The proposed model selection technique effectively determines the number of regression clusters.
• The developed computing procedure facilitates regression clustering estimation and selection.
• Simulation studies confirm the procedure's effectiveness.

Conclusions:

• The presented methods enhance the application of regression clustering.
• The approach is validated through simulations and a neuroscience dataset.
• This work provides a practical framework for regression clustering analysis.