A process to analyze data objects without consulting a known class label. The objects are clustered or grouped based on the principle of maximizing the intra-class similarity and minimizing the interclass similarity