Using AutoGluon for tabular data

In the previous chapter, we used Autopilot to see an example AutoML experiment that applies to the ACME Fishing Logistics use case. In this example, we are going to reproduce this experiment with AutoGluon. So, let's see how we can use AutoGluon to automate this task.

Note

The AutoGluon Tabular library benefits from compute instances with as much memory as possible. It is, therefore, recommended that AWS M5 instances (https://aws.amazon.com/ec2/instance-types/m5/) are used for tabular experiments. We will be using an m5.xlarge instance in this example and, therefore, running the example will incur AWS resource costs.

Prerequisites

Before we begin, there are a few fundamental topics that need to be accounted for, namely:

• At the time of writing, the AutoGluon library is not natively included as one of SageMaker's built-in estimators. This means that we will have to create our own Docker container for AutoGluon, using the...