Replication Data for: AdGazer: Improving Contextual Advertising with Theory-Informed Machine Learning

This repository provides the codes to reproduce the results from Jianping Ye, Michel Wedel, Pieters, AdGazer: Improving Contextual Advertising with Theory-Informed Machine Learning. We propose an Alternative Disclosure Plan for a delay of the release of our main dataset of the ad-context pairs. However, to maximize the reproducibility of the paper's results in this situation, we provide: 1. All preprocessed data used in the paper for model training, i.e. all features extracted from ad and context images in our main dataset with our algorithms; 2. All codes for the algorithms and models used in the paper; 3. All trained models (Sentence Transformer, XGBoost, CNN) and pre-trained models used; 4. Complete codes and data that reproduce the results for the study of Out-of-Distribution generalization, the study of the model interpretation and the study of the controlled experiment on ad placement; 5. Complete codes and partial data that illustrate how we conduct the study of the In-Silico experiments; 6. Complete codes for deploying the web app.