Clustering high‐dimensional data via feature selection