VCC-BPS: Vertical Collaborative Clustering using Bit Plane Slicing.

Area of Science:

• Data Science
• Machine Learning
• Clustering Algorithms

Background:

• Existing methods for uncovering data similarities across multiple sites are complex and prone to bias.
• Data owners need to analyze shared patterns without compromising data privacy.

Purpose of the Study:

• To propose a novel Vertical Collaborative Clustering using Bit Plane Slicing (VCC-BPS) method.
• To enhance accuracy and simplify collaboration in analyzing distributed datasets.
• To enable informed decision-making by identifying hidden data structures securely.

Main Methods:

• VCC-BPS transforms data into a code space, focusing on specific bit planes to capture local and collaborative similarities.
• The method identifies significant bits crucial for accurate classification and pattern recognition.
• Data compression is achieved by representing observations with fewer data symbols.

Main Results:

• VCC-BPS demonstrates improved accuracy, outperforming existing solutions on Geyser, Skin, and Iris datasets.
• The method achieves higher purity and a better Davies-Boulding index in collaborative clustering.
• Effective data compression capabilities were observed.

Conclusions:

• VCC-BPS offers a simple, accurate, and unique approach to vertical collaborative clustering.
• The method facilitates secure data analysis and decision-making across distributed data sites.
• VCC-BPS provides a valuable tool for collaborative pattern discovery and data compression.