Clustering Methods

Flashcards covering the key terms and concepts related to clustering methods discussed in the lecture.

Dendrogram

A tree diagram used to express relationships often in clustering.

Clustering

A method of finding natural groupings to see how things relate to each other. Samples within the group are more similar to each other than samples from different groups.

Supervised Method

Discriminant function analysis where groups are predefined.

Unsupervised Method

Clustering analysis where the data finds the groups.

Agglomerative

A clustering approach that builds up by adding sites or organisms to form a tree.

Divisive

A clustering approach that starts with one big group and chops it down to form groups.

Hierarchical Clustering

A method that proceeds step-by-step, where once an item is in a group, it stays in that group.

Non-Hierarchical Clustering

A method that allows items to switch groups during the process based on an iterative measure.

UPGMA

Unweighted paired groups method using arithmetic averages; a commonly used method in hierarchical agglomerative cluster analysis.

Single Linkage

Defines clusters by the smallest dissimilarity; can produce elongated dendrograms and cluster chains.

Complete Linkage

Defines clusters using the largest dissimilarities; sensitive to outliers.

Proportional Averaging

A method used in unweighted pairing to calculate average distances between clusters, dividing by the number of items in the clusters being compared.

K-means Clustering

A non-hierarchical method where K represents the number of clusters defined beforehand.

Elbow Plot

A method used to determine the optimum number of clusters by plotting the weighted sum of squares and looking for the 'elbow' in the plot.

Kalinsky-Harabasz Criteria

A method using the ratio of between-cluster variance to within-cluster variance to determine the number of clusters.