Category prediction resource

23 November 2015


The category prediction resource was created to help sellers and developers predict under which category a certain item should be listed. It’s currently working for Argentina, Brazil, Colombia, Chile and Mexico.


title The title of the item to predict. It must be a complete title in the language of the site. This parameter is mandatory.
category_from This parameter accepts a level 1 category and it’s used to limit the prediction to the subtree that spans from category_from as root. This parameter is optional.

Predicting by GET

Using the GET method it’s possible to predict one item at a time:


Predicting by POST

Using the POST method it’s possible to predict multiple (up to 10K) items at a time. When multiple items needs to be predicted, this method is recommended.

`curl -X POST -H "Content-Type: application/json" -d '{"title":"Ipod Touch Apple 16gb 5 Geração", "category_from":"MLB1743"}]' ""`

Response fields

idId of predicted category for the item.
nameName of the predicted category.
prediction_probabilityConfidence of the prediction. This value ranges between [0, 1], values closer to zero indicate low confidence whereas values closer to 1 indicate high confidence.
path_from_rootList with the categories from the root to the leaf (the predicted category). Every category is represented by ID and NAME attributes.
variationsthis field appears only when the predicted category supports variations.
shipping_modesShipping modes supported by the predicted category.

Learn how to build a good title so buyers can find your items and you can get the most accurate category prediction.