Clustering

NeuronDB provides multiple clustering algorithms for unsupervised learning.

K-Means

Partition data into k clusters:

Copy
CREATE TEMP TABLE kmeans_model AS
SELECT train_kmeans_model_id('data_table', 'features', 3, 100) AS model_id;

Mini-batch K-Means

Faster version for large datasets:

Copy
SELECT train_minibatch_kmeans('data_table', 'features', 3, 100) AS model_id;

DBSCAN

Density-based clustering:

Copy
SELECT train_dbscan('data_table', 'features', 0.5, 5) AS model_id;

GMM (Gaussian Mixture Model)

Probabilistic clustering:

Copy
CREATE TEMP TABLE gmm_model AS
SELECT train_gmm_model_id('data_table', 'features', 3) AS model_id;

Hierarchical Clustering

Copy
SELECT train_hierarchical_clustering('data_table', 'features', 3) AS model_id;

Learn More

For detailed documentation on clustering algorithms, choosing parameters, evaluating clusters, and visualization, visit:

Clustering Documentation

Related Topics

• Dimensionality Reduction - Reduce dimensions before clustering
• Quality Metrics - Evaluate clustering quality