Day 4 - Unsupervised/Anomaly Detection: RCF, PCA, IP Insights, Topic Models (LDA/NTM)

Today we consolidate unsupervised builtins that handle structure, anomaly, dimension, and topic without labels. Tests hint at these through signals "no labels / rare event / too many dimensions / discover topics." The key is distinguishing each algorithm's input/core parameters/common confusions.

Random Cut Forest (RCF) — Anomaly Detection

Create trees by cutting data randomly, assign high anomaly scores to points that are easy to isolate (rare).

• Use: Anomaly detection (unsupervised)