A differential evolution approach for simultaneous clustering and feature selection

In cluster analysis, the automatic evolution of clusters in the data without any prior assumption about the number of clusters is one of the most difficult process. Even, it is more difficult when feature selection is also under consideration. This paper introduces a differential evolution approach that simultaneously carries out clustering and feature selection processes. Our approach attempts to identify the cluster structure in the data automatically, while performing feature selection at the same time. To show the effectiveness of the approach, we analyze and compare it with traditional clustering approaches and recent simultaneous clustering and feature selection approaches on a variety of real-world benchmark datasets from UCI machine learning repository. The results indicate that our approach can better determine the cluster structure in the data than existing approaches, while using a reduced number of features.