Making batch predictions on Amazon ML is straightforward and follows this process:
- From the dashboard,
create a new Batch prediction. - Select the model.
- Select the datasource on which to apply the model.
- Set the prediction output folder and grant permissions.
- Review and launch.
We call the prediction dataset or datasource, the data on which we want to make predictions. In this chapter, we are in a testing context and the prediction dataset is the held-out dataset we extracted from the whole original dataset. In a real-world context, the prediction dataset refers to entirely new data and does not include the target variable.
The prediction can only work if the distribution of the prediction dataset is similar to the distribution of the training dataset on which the model has been trained. The prediction datasource and the training datasource must also share the same schema, with one difference the prediction dataset does not need to include the target variable. Amazon ML...