Summary

In this chapter, we have rounded out our advanced modeling topics by showing how to build H2O models in Spark pipelines with a hands-on sentiment analysis modeling example. We summarized the unsupervised learning methods available in H2O and showed how to build an anomaly detection model using the isolation forest algorithm for a credit card fraud transaction use case. We also reviewed how to update models, including refitting versus checkpointing, and showed requirements to ensure model reproducibility.

In Chapter 7, Understanding ML Models, we discuss approaches for understanding and reviewing our ML models.