Data mining - cluster analysis [2]

In this section:

1: Understanding Classification clustering [segmentation clustering, hierarchical clustering]

2: a method of dividing the cluster specific hierarchical clustering and

First, understand the classification clustering [segmentation clustering, hierarchical clustering]

Second, the specific method of clustering and hierarchical clustering segmentation

The cluster is divided

Given data set D n data objects, and to generate the number of clusters k, the data partitioning algorithm objects k (k <= n) partitions, wherein each partition represents a cluster. And k satisfy the following conditions:
1: Each cluster contains at least one object
2: Every object belongs to one and only one cluster.

Common methods:
k- means
k- center value (k-means for sensitive isolate the value of [averaging which maxima minima big impact In order to solve this we
It introduces the idea of ​​a central value)
　　The algorithm does not use the average value as a reference point in the cluster, the cluster can use the object as the center point, i.e., the most central point of reference.
　　It is now almost k- means method of calculating the difference: only the data points in the sample space can be used as a central point

Hierarchical clustering can be divided into two categories:

Agglomerative clustering:

Initially, each data point as a cluster

Each step merger nearest cluster, a cluster until far

Cluster division:

Initially, all data points are seen as a bunch of

Further dividing each of a cluster, each cluster contains only a known data points

• Conventional hierarchical clustering method using the similarity or distance matrix : a split or merge each cluster