[r] zhang96, "Birch: An Efficient Data Clustering Method for Very Large Databases"

@inproceedings{ zhang96birch,
  author = "Tian Zhang and Raghu Ramakrishnan and Miron Livny",
  title = "{BIRCH}: an efficient data clustering method for very large databases",
  pages = "103--114",
  year = "1996",
  url = "citeseer.ist.psu.edu/zhang96birch.html" }

Motivation:
Clustering in large databases is very expensive. The reason is due to excessive database scans. This paper addresses this problem.
Contributions:
Proposed the idea of "local clustering" and proved that introduction of "local clustering" didn't affect correctness. This idea is quite like B-Tree in database field, so every cluster caculation and update is done locally.
Method:
Proved that Cluster Feature (CF) had some interesting properties; Introduced the algorithms to deal with CFs.
Discussion:
The idea in this paper is influential. Recently, J. Han's group propsed moving objects clustering based on "microclusters". Their ideas are largely borrowed from Birch.