CDF Transform-and-Shift: An effective way to deal with datasets of inhomogeneous cluster densities

The problem of inhomogeneous cluster densities has been a long-standing issue for distance-based and density-based algorithms in clustering anomaly detection. These implicitly assume that all clusters have approximately the same density. As result, they often exhibit bias towards dense presence sparse clusters. Many remedies suggested; yet, we show are partial solutions which do not address satisfactorily. To match implicit assumption, propose to transform given dataset such transformed density while regions locally low become globally -- homogenising preserving structure dataset. We this can be achieved by using new multi-dimensional Cumulative Distribution Function transform-and-shift method. method applied every dataset, before is used many existing their assumption without algorithmic modification. proposed performs better than remedies.