CriteoAds¶

Criteo Display Advertising Challenge.

This is a 100,000-row sample of the Criteo Display Advertising Challenge dataset: real ad-impression logs where the goal is to predict whether an ad was clicked. Each row has a binary click target, 13 integer features (I1 to I13, mostly anonymised counts, frequently missing) and 26 categorical features (C1 to C26, hashed to opaque strings). The categorical fields are extremely high-cardinality, which makes this a natural fit for one-hot models such as linear_model.AdPredictor.

Integer features are returned as int (or None when missing) and categorical features as str (or None when missing).

Attributes¶

• desc

  Return the description from the docstring.

• is_downloaded

  Indicate whether or the data has been correctly downloaded.

• path

Methods¶

References¶

• Display Advertising Challenge - Criteo AI Lab

• Kaggle competition page