A COLLABORATIVE POSSIBILISTIC FUZZY C-MEANS CLUSTERING APPROACH FOR MULTI-DIMENSIONAL DATA ANALYSIS

Abstract

The rapid development of data acquisition technologies has led to an explosion of data sources. Many traditional data mining techniques and methods have become outdated and are no longer suitable for solving large, high-dimensional data problems. The paper proposes improving the collaborative possibilistic fuzzy clustering algorithm for multi-dimensional data analysis using random projection feature reduction. The random projection feature reduction technique allows for the preservation of relative distances after dimensional reduction, which can help reduce computational complexity while still ensuring the accuracy of the proposed algorithm compared to the algorithm before dimensionality reduction. The proposed algorithm implemented on the collaborative clustering model can help share information about cluster structure at data sites during computation, allowing problems to be performed where data is located on different computers in a network. Experiments performed on two multidimensional datasets downloaded from the UCI
Machine Learning Repository library and remote sensing image data show that the proposed method yields significantly better results than some previously proposed methods. These experimental results demonstrate the potential of developing collaborative clustering models, combined with dimensionality reduction techniques, to tackle high-dimensional and distributed large data problems.