UL methods in H2O
H2O includes several unsupervised learning algorithms including Generalized Low Rank Models (GLRM), Principal Component Analysis (PCA), and an aggregator for dimensionality reduction. Clustering use cases can utilize k-means clustering, H2O aggregator, GLRM, or PCA. Unsupervised learning also underlies a set of useful feature transformers used in predictive modeling applications—for example, the distance of an observation to a specific data cluster identified by an unsupervised method. In addition, H2O provides an isolation forest algorithm for anomaly detection.
What is anomaly detection?
Most machine learning (ML) algorithms attempt, in some manner, to find patterns in data. These patterns are leveraged to make predictions in supervised learning models. Many unsupervised learning algorithms try to uncover patterns through clustering similar data or estimating boundaries between data segments. Unsupervised anomaly detection algorithms take the opposite...
