- Ali-CCP：Alibaba Click and Conversion Prediction
- Papper - Entire Space Multi-Task Model: An Effective Approach for Estimating Post-Click Conversion Rate
Based on the given Description of feature sets table
- Find similar (exact same (because of the size of dataset is too big)) user by User Features
- Using this user info to build a prediction engine
- In the test set, we assume we build a recommend mechanism that we can use a pop-up window to force user in test set to click it.
- If in the test set the user actually click and buy that kind of goods. Then we say it's a good prediction (recommendation)