- 최신
- 최다 투표
- 가장 많은 댓글
Hi & thanks for asking!
As mentioned in the "Metadata fields" section of the Custom Datasets and Schemas developer guide page:
If you add your own metadata field of type string, it must include the
categorical
attribute or thetextual
attribute (only Items schemas support fields with the textual attribute). Otherwise, Amazon Personalize won't use the field when training a model.
...So you must explicitly specify string fields as either categorical or textual, otherwise Personalize will ignore them in model training (which can also be useful - if you'd like to include a field for filtering purpose but not have it affect the model). Since you're building a user metadata schema, not item - it'll need to be categorical.
However - I think you'll also need to remove the categorical
flag from your "AGE" field there: since it's a numeric rather than string. You can also refer to this example user metadata schema.
To expand on what Alex said, numerical fields are always used model training (for the recipes and recommenders that use that data). You can have the AGE field as categorical field if you want to define your own categories, for instance "child", "teen", "adult". You should do this if you have particular categories that make sense for your use-case rather than relying on Amazon Personalize to process the data. Non categorical/non numerical fields can be added to be used for filtering and numerical fields can be used for optimizing on an additional objective.