Group Records Dynamically with Unsupervised Clustering

Unsupervised clustering groups similar records using pre-trained algorithms without training or maintaining a reusable model, which is useful for exploring data, collapsing several columns into one variable, or summarizing millions of rows into a few clusters for downstream models. Choose K-Means to split records into a set number of clusters, or HDBSCAN to find natural groupings and outliers without specifying a cluster count, then run a predict job to get cluster IDs and labels. For reusable assignments and high-volume scoring across standard objects, use structured clustering instead.