[KDD-2018] KDD Cup Call for Participation: Fresh Air

[Ping Zhang]

Website: https://biendata.com/competition/kdd_2018/

**********
Timeline
**********
2018/3/15: Release historical data of Beijing between 2017/1/31 and 2018/1/31. You can start exploring and constructing your forecasting framework using this dataset.
2018/3/31: Release historical data of the second city. The generalization of your forecasting framework will be challenged on both cities in this competition.
2018/4/1 – 2018/4/30: Phase I: Practice Only. This is the public testing stage. The scores you get in this stage will not be counted in the final ranking. This is the time period that you explore the modeling framework, and develop and stabilize your data downloading, model retraining, forecasting, and submission process programmatically.
2018/5/1 0:00 a.m. (UTC) – 2018/5/31 23:59 p.m. (UTC): Phase II: Live Testing and Final Stage. The results of this phase will decide the final leaderboard of this year's KDD Cup.
2018/5/7 23:59 (UTC): Team merger deadline: you cannot create or change team after the team merger deadline.

**********
Prize
**********
The amount of all prizes is $21,500.
Gold Medal (1st Place): $10,000
Silver Medal (2nd Place): $5,000
Bronze Medal (3rd Place): $3,000
Honorable Prizes (4th – 10th Place): $500, each team.
More prizes, such as travel fund, certificate, or additional cash prize remain to be announced.

**********
Background
**********
Over the past years, air pollution has become progressively more severe in many large cities, such as Beijing. In 2011, an article in the Los Angeles Times cited Dane Westerdahl, an air quality expert from Cornell University, describing the air quality of Beijing as 'downwind from a forest fire'. Among different air pollutants, air particles, or Particulate Matters (PM), are one of the deadliest forms. Particles with a diameter of 2.5 μm or less (called PM2.5) can penetrate deeply into human lungs and enter blood vessels, causing DNA mutations, cancer, central neural system damage, and premature death.

Existing biomedical research demonstrates that, once inhaled, PM2.5 can hardly be self-cleaned by the human immune system. Therefore, accurately monitoring and predicting the concentration of PM2.5 and other air particles have become increasingly crucial. With precise predictions of air pollution levels, the public and governments can respond with appropriate decisions, such as closing schools and discouraging outdoor activities, to greatly mitigate the harmful consequences of air pollution.

********************
Task and Evaluation
********************
Participants are requested to predict PM2.5, PM10, and O3 concentration levels over the coming 48 hours for two cities, one of which is Beijing, China while the other will be revealed on March 31. On each day throughout the competition, air quality data and meteorological data for both cities will be provided on the hourly basis. For example, on May 14, the participants will be able to access historical data up to May 14 (including), and will have to predict the pollution level for May 15 and 16. Over a period of 24 hours (by 23:59 UTC), each team will be allowed to make no more than 3 submissions to predict 48 hours of air quality results, starting from 0:00 UTC of the next day. We will soon provide more details on the submission API and the submission file format.