jd-opensource / cim

Implementaion of the WWW paper Implicit User Awareness Modeling via Candidate Items for CTR Prediction in Search Ads

Implicit User Awareness Modeling via Candidate Items for CTR Prediction in Search Ads

This repo is the official implementation for the WWW 2022 paper: Implicit User Awareness Modeling via Candidate Items for CTR Prediction in Search Ads.

Data format

Our released dataset CandiCTR-Pub could be found at JD JingPan with password xngmgr.

In the data files, each row corresponds to a search session. Each column in the data represents userID, queryID, label list, target items and request item queue. Each item is consist of itemID, categoryID, brandID, vendorID and priceID. All data have been desensitized.

Requirements

• python 3.6.13
• tensorflow 1.15.0
• scikit-learn 0.24.2

Quick start

Create a new data folder and put the downloaded dataset into the folder. Then,

python src/main.py