Baseline Approaches for App Advertising Conversion Rate Prediction Competition

The competition on app advertising conversion rate prediction has attracted over 16,000 registrations, with thousands of teams submitting results. Participants are asked to estimate the probability that a user activates an app after clicking an ad, a classic binary classification problem similar to pCTR prediction.

The provided dataset includes user profiles, ad click logs, app installation records, and related streams, offering ample opportunities for feature engineering and model development while presenting several challenges.

Two baseline solutions are released on GitHub. Version 1 predicts conversion rates by using historical average rates of fields such as creativeID, adID, and appID from ad.csv . Version 2 applies simple one‑hot encoding to the same fields and trains a Logistic Regression model.

These baselines illustrate how to preprocess data, encode features, train models, and generate submission files in Python, highlighting the potential of machine‑learning methods over simple statistical approaches.

For more details, participants can visit the official competition website and follow the provided links to register and access additional resources.