Density-based algorithms have attracted many researchers due to their ability to identify clusters with arbitrary shapes in noisy datasets. DENCLUE is a density-based algorithm that clusters objects based on a density function instead of proximity measurements within data. DENCLUE is efficient in clustering high-dimensional datasets. However, it has difficulty in discovering clusters with highly varying densities. To overcome this issue, this study proposes an enhanced variant of the DENCLUE algorithm, called VDENCLUE, based on the varying Kernel Density Estimation. The VDENCLUE uses the local features of the data space, so clusters with arbitrary shapes and densities can be identified. In order to demonstrate the effectiveness of its approach, VDENCLUE was empirically evaluated and compared to the DENCLUE algorithm. Experimental results show that in almost all datasets, the VDENCLUE algorithm outperforms the DENCLUE algorithm.