In this final chapter, we will discuss a few approaches to tuning models. We will cover ways of addressing missing data. Although we have used example datasets without any missing data, in the real world missing data is a common occurrence. We will also discuss what can be done when a model is performing poorly, including a detailed examination of how to search for and optimize model hyperparameters.
This chapter will cover the following topics:
Dealing with missing data
Solutions for models with low accuracy
In this chapter, we make use of two new packages: the gridExtra package for graphics and the mgcv package for fitting generalized additive models at the end. These new packages should be added to the checkpoint.R
file, and the file should be sourced to set up the R environment for the rest of the code shown. R can be set up and an H2O cluster initialized using the following code:
source("checkpoint.R") options(width = 70, digits = 2) cl <- h2o...