Dataset description:
The Criteo dataset is a widely-used benchmark dataset for CTR prediction, which contains about one week of click-through data for display advertising. It has 13 numerical feature fields and 26 categorical feature fields. Following the setting with the AutoInt work, we randomly split the data into 8:1:1 as the training set, validation set, and test set, respectively.
The dataset statistics are summarized as follows:
Dataset Split
Total
#Train… See the full description on the dataset page:
https://huggingface.co/datasets/reczoo/Criteo_x4.