An Improvement of Fuzzy C-Means Clustering Using the Multiple Kernels Technique with Gravitational Force Information for Data Classification

It can be challenging for clustering algorithms to handle data that is noisy, multidimensional, and overlapping. A new approach has been presented in this paper for fuzzy c-means clustering using the technique of multiple kernels with gravitational force information for data classification (GF-MKFCM). This method is based on the gravitational physical theory, which suggests that data samples with high density will attract data samples with low density towards them. This means that data samples with higher density will be closer to the cluster center. Additionally, to reduce data overlap, the data is represented in multiple kernels space, making it easier to cluster. We tested the proposed method using two types of kernel functions: Polynomial and Gaussian. Experimental results on data sets from the UCI machine learning library and remote sensing image data show that the accuracy of the proposed algorithm is significantly higher than some other algorithms. © 2023 IEEE.