CLUSTERING USING BIRCH || DATA MINING || MACHINE LEARNING -Malayalam || KTU
Oct 5, 2023
Hi all, today we are going to discuss about BIRCH ALGORITHM
So, what is BIRCH?
BIRCH is Balanced Iterative Reducing and Clustering using Hierarchies
That is BIRCH
What is BIRCH?
Balanced Iterative Reducing and Clustering using Hierarchies
In short, what is BIRCH?
It is designed for clustering a large amount of numeric data by interchanging hierarchical clustering
Where is BIRCH used?
BIRCH algorithm is designed for clustering a large amount of numeric data
So, how does BIRCH work?
BIRCH algorithm integrates hierarchical clustering and other clustering methods like Iterative Partitioning
BIRCH overcomes two difficulties in agglomerative clustering methods
Scalability and Inability to undo what was done in the previous step
There are some notations used in BIRCH
Clustering Feature used to summarize a cluster
Clustering Feature Tree used to represent cluster hierarchy
What is Clustering Feature?
Clustering Feature of a Cluster is a 3D vector summarizing information about clusters of objects
Clustering Feature of a Cluster is denoted by CF
CF is equal to N Ls Ss
N is number of points
Ls is linear sum of n points
Linear sum is sigma i is equal to 1 to n xi
Ss is square sum of data points
sigma i is equal to 1 to n xi
Clustering Feature is a 3D vector summarizing information about clusters of objects
Clustering Feature of a Cluster is a 3D vector summarizing information about clusters of objects
Clustering Feature is a summary of statistics for a given cluster
Clustering Feature of a Cluster is a 3D vector summarizing information about clusters of objects
Radius is average distance from the member of objects to the centroid
D is average pairwise distance within the cluster
Clustering Feature of a Cluster is a 3D vector summarizing information about clusters of objects
Suppose there are 3 points in a cluster C1
Clustering Feature of a Cluster is a 3D vector summarizing information about clusters of objects
Suppose there are 3 points in a cluster C1
Suppose there are 3 points in a cluster C2
Clustering Feature of a Cluster is a 3D vector summarizing information about clusters of objects
Clustering Feature of a Cluster is a High-Balanced Tree that stores the clustering features for the hierarchical clustering
Phase-1 scans the database to build an in-shell in-the-memory CF Tree, which can be viewed as a multilevel compression of data that tries to preserve the data's inherent clustering structure
Phase-2 scans the database to build an in-shell in-the-memory CF Tree, which can be viewed as a multilevel compression of data that tries to preserve the data's inherent clustering structure
Phase-1 scans the database to build an in-shell in-the-memory CF Tree, which can be viewed as a multilevel compression of data that tries to preserve the data's inherent clustering structure
An object is inserted into the closest leaf entry or subcluster. If the diameter of the subcluster stored in the leaf node after insertion is larger than the threshold value, then the leaf node and possibly the other nodes are split.
After the insertion of a new object, information about the object is passed to the root of the tree.
If the size of the memory that is needed for storing the CF Tree is larger than the size of the main memory, then a larger threshold value can be specified and the CF Tree is rebuilt.
Input D is equal to T1, T2, etc. up to Tn.
T1 is the threshold for CF Tree construction.
Phase 2 uses an arbitrary clustering algorithm to cluster the leaf nodes of the CF Tree. Also, we can use any algorithm such as partitional algorithm can be used in phase 2.
Phase 2 removes the sparse clusters as outliers and groups dense clusters into larger ones. The time complexity of the bridge algorithm is O of n, where n is the number of objects to be clustered.
What are the advantages and disadvantages of the bridge algorithm?
It scales linearly, finds a good clustering with single scan, and improves the quality with fewer additional scans.
It fixes the size of the leaf nodes, so clusters may not be so natural.
It handles only numeric data and sensitive to insertion order of data points.
So these are the advantages and disadvantages of bridge algorithm.

