Interactive K-Means Clustering Method Based on User Behavior for Different Analysis Target in Medicine

Area of Science:

• Data Science
• Medical Informatics
• Machine Learning

Background:

• Clustering algorithms are vital for data analysis but struggle with high-dimensional data, particularly in medicine.
• Standard clustering may overlook crucial business relationships, leading to results misaligned with user objectives.
• Integrating user knowledge, such as physician expertise, is key to improving clustering relevance.

Purpose of the Study:

• To propose an interactive K-means clustering method that enhances user satisfaction with analysis results.
• To address the limitations of traditional clustering in high-dimensional medical data.
• To align clustering outcomes with specific user business goals and analytical intents.

Main Methods:

• An interactive K-means clustering approach incorporating user feedback for iterative refinement.
• Utilizing particle swarm optimization to tune algorithm parameters, focusing on weight settings.
• Optimizing parameters to reflect user business preferences and analytical requirements.
• Validation using a breast cancer dataset to demonstrate practical application.

Main Results:

• The proposed interactive method significantly improves user satisfaction with clustering outcomes.
• Parameter optimization via particle swarm optimization effectively incorporates user preferences.
• The method demonstrates superior performance compared to standard clustering techniques in the breast cancer case study.
• Enhanced clustering results are more aligned with user-defined business goals.

Conclusions:

• Interactive K-means clustering, enhanced by user feedback and particle swarm optimization, offers a powerful solution for medical data analysis.
• This approach effectively bridges the gap between algorithmic results and user-specific business needs.
• The method shows promise for improving the utility and relevance of clustering in clinical and research settings.