This repository includes experiment code for two risk-aware bidding strategies.

For repeating the experiment, please follow these steps:

i) Download iPinYou datasets from http://data.computational-advertising.org/.

ii) Prepare data following https://github.com/wnzhang/make-ipinyou-data.

iii) Use remove-pinyou-tag.py to remove tag information from datasets.

iv) Use lryzx.py to train LR models for each campaign.

v) Use divide-pinyou-test-data.py to split test data into two equal parts.

vi) Use lap-bayes.py to train CTR distribution model.

vii) Use replay-lr.py, replay-rmr.py, replay-var.py to replay validation test with the baseline and the two risk-aware strategies.

viii) Use select-parameters.py to select model hyperparameters from replay results.

ix) Use experiment.py to replay test data and obtain final results.