Fuzzy c-means clustering of incomplete data

Area of Science:

• Data Science
• Machine Learning
• Statistics

Background:

• Clustering algorithms like fuzzy c-means (FCM) are vital for data analysis.
• Standard FCM requires complete data, limiting its application to datasets with missing values.
• Incomplete data is common in real-world scenarios across various scientific domains.

Purpose of the Study:

• To develop and evaluate methods for applying FCM to incomplete datasets.
• To address the challenge of clustering data with missing feature values.
• To enhance the applicability of FCM in practical data mining scenarios.

Main Methods:

• Introduced four distinct strategies for FCM clustering of incomplete data.
• Three strategies involved modified versions of the fuzzy c-means algorithm.
• Numerical convergence properties of the proposed algorithms were analyzed.

Main Results:

• Successfully adapted FCM for datasets containing missing feature values.
• Evaluated the performance of the new algorithms using both real and synthetic incomplete data.
• Demonstrated the feasibility and effectiveness of the proposed clustering strategies.

Conclusions:

• The developed strategies enable effective fuzzy c-means clustering on incomplete datasets.
• These methods expand the utility of FCM for real-world data analysis challenges.
• Further research can explore the convergence and performance in more complex scenarios.