An Improved Fast Search Clustering Algorithm Based on Kernel Density

摘要

Clustering is an important algorithm for data mining. FSC is a kind of clustering algorithm based on density, which has been proposed in the journal Science in 2014. FSC only requires one input parameter and has a higher practicability. RFSC, which is an improved algorithm of FSC algorithm, is less sensitive to the input parameters and faster. However, neither RFSC nor FSC can deal with uneven density data sets. In order to solve that problem, we propose an improved algorithm KFSC in this paper by dynamically controlling of the width of the kernel function. KFSC uses the idea of attractor of the DENCLUE and can customize their own personalized attraction for each point. The experimental results on synthetic data sets show that KFSC has a better performance on uneven density data sets than FSC and RFSC.