Most clustering algorithms are running in static conditions

which means that it does not allow any incremental data to be added. An incremental clustering algorithm based on K-nearest neighbor（ KNN） is proposed. This algorithm has two innovations

one is using KNN idea and Sample Compactness to deal with incremental data. The other is splitting or merging cluster according to change of cluster feature. Experiments show that the algorithm can not only find new clusters

but also can effectively avoid the noise points

and can deal with non-spherical datasets.