Problem Statement

A leading affiliate network company from Europe wants to leverage machine learning to improve (optimise) their conversion rates and eventually their topline. Their network is spread across multiple countries in europe such as Portugal, Germany, France, Austria, Switzerland etc.

Affiliate network is a form of online marketing channel where an intermediary promotes products / services and earns commission based on conversions (click or sign up). The benefit companies sees in using such affiliate channels is that, they are able to reach to audience which doesn’t exist in their marketing reach.

The company wants to improve their CPC (cost per click) performance. A future insight about an ad performance will give them enough headstart to make changes (if necessary) in their upcoming CPC campaigns.

In this challenge, you have to predict the probability whether an ad will get clicked or not.

Download Dataset

Data Description

You are given three files to download: train.csv, test.csv and sample_submission.csv Variables in this data set are anonymized due to privacy.
The training data is given for 10 days ( 10 Jan 2017 to 20 Jan 2017). The test data is given for next 3 days.

Variable	Description
ID	Unique ID
datetime	timestamp
siteid	website id
offerid	offer id (commission based offers)
category	offer category
merchant	seller ID
countrycode	country where affiliates reach is present
browserid	browser used
devid	device used
click	target variable

Submission

A participant has to submit a zip file containing your ID and predicted probabilities in a csv format. Check the sample submission file for format.

ID,click
IDE4beP,0.125
IDfo26Y,0.015
IDYZM6I,0.201
ID8CVw1,0.894
IDPltMK,0.157

Evaluation Metric

Submission will be evaluated based on AUC-ROC score. Higher the better. To know more, read here.

Scripts

LightGBM R (0.63 LB) - Click Here
LightGBM Python (0.67 LB) - Click Here
CatBoost in Python (0.675 LB) - Click Here
Stacking Starter in R - Click Here
Keras Starter in Python - Click Here

Update

[Aug 09, 16:00] Given that 5 days are left and some participants are about to reach limit, the maximum submission limit has been increased to 200, earlier it was 100.
[Aug 21, 18:10] The Private Leaderboard is Live. We'll reach out to winners and soon share the winning solution.
[Aug 22, 19:18] Sincere apologies for this inconvenience. We've discovered that private LB scores aren't updated for all participants, this is due to our backend issue. Therefore, please do not consider the current leaderboard as final. We'll update you when it is done.
[Aug 28, 17:00] Private Leaderboard is Live Now. You can check your final rankings.