Summary

In this chapter, we have considered different options for splitting data, explored, in some depth, powerful and popular algorithms such as gradient boosting and random forest, learned how to optimize model hyperparameters using a two-stage grid search strategy, utilized AutoML to efficiently fit multiple models, and further investigated options for feature engineering, including a deep dive into target encoding. Additionally, we saw how Flow can be used to monitor the H2O system and investigate data and models interactively. You now have most of the tools required to build effective enterprise-scale predictive models using the H2O platform.

However, we are not finished with our advanced modeling topics. In Chapter 6, Advanced Model Building – Part II, we will discuss best practices for data acquisition, look in more depth at checkpointing and refitting models, and show you how to ensure reproducibility. Additionally, we will thoroughly consider two more hands-on examples...