Clusterability Detection and Initial Seed Selection in Large Data Sets

摘要: The need for a preliminary assessment of the clustering tendency or clusterability massive data sets is known. A good detection method should serve to in uence decision as whether cluster at all, well provide useful seed input chosen algorithm. We present framework de nition set from distance-based perspective. discuss graphbased system detecting and generating information including an estimate value k { number clusters set, parameter many methods. output our tunable accommodate wide variety have conducted experiments using methodology with stock market well-known BIRCH sets, two higher dimensions. Based on results we nd that can basis much future work this area. report promising directions.