Quantum jump clustering

Data transformation is an important aspect of cluster analysis. Data normalization and feature weighting are two examples of data transformation where normal feature space (original data) is converted into transformed feature space. Data transformation can help to produce better clustering results and extract meaningful information/rules. In this paper we propose a new transformation technique inspired by quantum jumps using Bohr´s hydrogen model. Feature weighting is incorporated into a quantum jump algorithm to obtain a transformed feature space that leads to better groupings (clusters). The algorithm is tested on simulated and real world datasets. The results demonstrate the feasibility of this algorithm for datasets that are known to cause problems to standard clustering algorithms.