We have now created several data sources based on the original Titanic dataset in S3. We are ready to train and evaluate an Amazon ML prediction model. In Amazon ML, creating a model consists of the following:
- Selecting the training datasource
- Defining a recipe for data transformation
- Setting the parameters of the learning algorithm
- Evaluating the quality of the model
In this chapter, we will start by exploring the data transformations available in Amazon ML, and we will compare different recipes for the Titanic dataset. Amazon ML defines recipes by default depending on the nature of the data. We will investigate and challenge these default transformations.
The model-building step is simple enough, and we will spend some time examining the available parameters. The model evaluation is where everything converges. The evaluation metrics are dependent on the type of the prediction at hand, regression, binary or multi-class classification. We will look at how these different...