Evolving Cauchy Possibilistic Clustering for Data Stream Modeling and Classification

This article proposes a novel approach to clustering called eCauchy, which is based on the Cauchy density function. This method is similar to the well-known possibilistic c-means clustering (PCM), which was developed to address the limitations of fuzzy c-means algorithms (FCM). Density-based clustering is particularly useful when dealing with noisy data and frequent outliers. The recursive measure of density can be adapted for various data mining tasks, including regression and classification. The eCauchy algorithm is simple to use and can handle big-data problems. It evolves dynamically by adding and removing clusters based on a few initial parameters, such as maximum and minimum density. This allows for the identification of clusters with different shapes and sizes and is robust to outliers and high levels of noise. The paper demonstrates the universality of this algorithm and its modifications through several simple examples that can also be found in more complex forms in real-world applications.