Associations feature supports the following fields:
- Numeric: it will be discretized in segments to be transformed into categorical.
- Categorical: each category will be considered a different value.
- Text: each term will be considered a different value. You can change this behaviour from the configuration source panel by tuning the text analysis parameters.
- Items: each item will be considered a different value. BigML automatically detects items fields when they are separated by non-alphanumeric characters, however any single character can be a separator. These kinds of fields are usually found in transactional datasets containing a set of items. For example, for market basket analysis you can find transactions like: "skimmed milk, dark chocolate, carrots". You will have to set a unique separator for your items so BigML can automatically recognize the unique items (in the previous example the separator is the comma).
Please read the sources chapter of the BigML Dashboard documentation to learn more about the field types, or refer to the associations documentation to learn how to prepare your data for associations.