Sparse Regularization in Fuzzy c-Means for High-Dimensional Data Clustering.

Area of Science:

• Data Science
• Machine Learning
• Statistics

Background:

• Clustering high-dimensional data is challenging due to the curse of dimensionality.
• Identifying relevant features is crucial for accurate cluster discovery.
• Existing methods struggle to simultaneously select features and determine clusters effectively.

Purpose of the Study:

• To propose a novel fuzzy c-means (FCM) model incorporating sparse regularization.
• To address the challenge of identifying relevant features in high-dimensional data clustering.
• To develop an efficient algorithm for solving the proposed model.

Main Methods:

• Reformulated the FCM objective function into a weighted between-cluster sum of squares form.
• Introduced ℓq (0
• Developed an explicit algorithm for model optimization, with analytic solutions for q=1 and q=1/2.

Main Results:

• The proposed model effectively shrinks weights of irrelevant (noisy) features to zero.
• Demonstrated superior performance compared to existing clustering models on synthetic and real-world datasets.
• Achieved efficient solutions, particularly in analytic forms for specific regularization parameters.

Conclusions:

• The novel sparse regularized FCM model offers an effective solution for feature selection and clustering in high-dimensional data.
• The method enhances clustering accuracy by focusing on relevant features.
• The developed algorithm provides an efficient and robust approach for practical applications.