CS 312 AI Clustering Algorithms

An algorithm that groups data points into clusters based on minimizing distances to cluster centers.

Groups of data points with high similarity within the group.

Data points representing the center of a cluster.

Two-step process of assigning data points to clusters then recalculating cluster centers.

How spread out the data points are inside a cluster.

Sum of Squared Errors, measure of error in k-means clustering.

The number of clusters to create.

Straight-line distances between data points.

A technique used in k-means clustering to determine the optimal number of clusters by finding the 'elbow point' on a graph of SSE (Sum of Squared Errors) against the number of clusters.

A measure of the total error in k-means clustering. It calculates the sum of squared distances between each data point and its assigned cluster centroid.

A metric used to evaluate cluster quality by measuring how well each data point fits into its assigned cluster compared to other clusters.

The degree to which data points within a cluster are similar or tightly grouped together.

The degree to which different clusters are distinct or well-separated from each other.

The ideal number of clusters that balances the need for minimal error (good fit) and a reasonable number of clusters for interpretability.

The balance between minimizing errors in k-means clustering and keeping the number of clusters manageable for understanding the data.

The number of clusters determined by looking for the 'elbow point' in the graph, where the SSE curve begins to flatten.

A method that groups data points into a tree-like structure based on their similarity. It can use either a 'bottom-up' or 'top-down' approach to form clusters.

A 'bottom-up' approach to hierarchical clustering where data points are initially separated and gradually merged into larger clusters based on their similarity.

A 'top-down' approach to hierarchical clustering where the entire dataset is initially a single cluster and is progressively split into smaller clusters based on dissimilarity.

A tree-like diagram representing hierarchical clustering structure. It shows how clusters are formed and their relationships.

A dendrogram visually represents the hierarchical clustering results, showing the relationships and formation of clusters at different levels. It helps analyze the data's inherent structure and identify optimal groupings by cutting the tree at specified depth.

By cutting the dendrogram at a specific level, you get k groups of smaller dendrograms, which represent the final clusters. The cut level determines the number of clusters and their composition.

A high silhouette coefficient indicates that data points are well-clustered, meaning they are more similar to members of their own cluster than to members of other clusters.

Divisions are based on similarity (agglomerative) or dissimilarity (divisive) between data points. Algorithms like Ward's method or single linkage are used to measure these distances and determine the optimal groupings at each level of the hierarchy.