eFCM: An Enhanced Fuzzy C-Means Algorithm for Longitudinal Intervention Data

Area of Science:

• Data Science
• Machine Learning
• Biostatistics

Background:

• Clustering methods are crucial for analyzing heterogeneity in treatment effects, particularly in longitudinal behavioral intervention studies.
• Existing methods like K-means and Fuzzy C-means (FCM) are widely used but face challenges with high-dimensional data, missing values, and initialization issues.
• The MIFuzzy framework aims to address these methodological challenges in analyzing longitudinal intervention data.

Purpose of the Study:

• To propose a novel initialization method for Fuzzy C-means (FCM) to overcome local optima and reduce convergence time.
• To enhance the MIFuzzy framework by incorporating an improved FCM algorithm (eFCM) for high-dimensional longitudinal intervention data with missing values.
• To address overlapping clusters and improve the analysis of complex datasets.

Main Methods:

• Developed an enhanced Fuzzy C-means clustering (eFCM) algorithm inspired by K-means++ initialization.
• Integrated eFCM into the existing MIFuzzy framework to handle high-dimensional longitudinal data with missing values.
• Evaluated the proposed method on real-world longitudinal intervention data and standard benchmark datasets.

Main Results:

• The enhanced Fuzzy C-means (eFCM) method demonstrates improved computational efficiency compared to conventional FCM.
• eFCM effectively avoids the problem of local optimization, leading to more robust clustering results.
• The method successfully handles high-dimensional longitudinal data with missing values and overlapping clusters.

Conclusions:

• The proposed eFCM integrated into MIFuzzy offers a significant advancement for analyzing complex longitudinal intervention data.
• This approach enhances the reliability and efficiency of clustering in behavioral intervention studies with missing data.
• The findings suggest eFCM is a valuable tool for identifying distinct patient groups and understanding treatment effect heterogeneity.