[hpc-announce] Call for Data Mining Competition -- FedCSIS at the KnowledgePit Platform

Marcin Paprzycki marcin at amu.edu.pl
Sat Nov 30 16:19:15 CST 2019

We invite industrial, government, and academic organizations to submit 
proposals to organize a Data Mining Competition at the 15th Conference 
on Computer Science and Information Systems (FedCSIS 2020). Data Mining 
Competitions are based on data sets released by the sponsoring 
organizations. The competitions will last for several months, with the 
winners announced at the conference on September 6-9, 2020.

Data Mining Competitions associated with the FedCSIS Conference Series 
have a long tradition. Since 2014, they have been organized every year 
at the KnowledgePit Platform, and in total, they attracted over 1200 
teams from more than 40 countries. The competitions are usually focused 
on learning the most accurate classification / prediction / forecasting 
models based on the provided data sets. The participants are free to use 
all machine learning / data mining methods that they find as useful.

Those who are willing to organize a competition need to remember about 
the following:

The tasks should be challenging but achievable within the expected 
timeline of the competition.

The data sets should be novel. They cannot be available before the 
competition begins. The KnowledgePit Team will assist in preparation of 
publicly available versions of the competition data sets.

The organizers will be requested to cooperate with the KnowledgePit Team 
in order to describe the task and the data in the most clear way to the 
competition participants.

We will consider topics from (but not limited to) the following 
application domains:

- Mobility
- Health and wellness
- Finance
- Advanced manufacturing
- Predictive maintenance
- Retail and e-commerce
- Gaming industry
- Cybersecurity
- Smart City and Community

The proposals will be evaluated based on the significance and novelty of 
the considered problem, clarity of its definition, and availability of 
the related data. The competition will be hosted at the KnowledgePit 
Platform which provides an automatic submission and evaluation system, 
and maintains a leaderboard for the participants. Competition winners 
will be selected based on their final ranking, as well as a short report 
describing their algorithms and results. Top performing teams will also 
be invited to publish a regular paper at FedCSIS 2020, where they will 
be able to present their solution during the special competition 
session. The additional paper will be prepared and presented jointly by 
the Competition Organizers and the KnowledgePit Team.

Tentative Competition Timeline.

Step 1 – Submissions of competition proposals due, December 20, 2019

Step 2 – Proposal evaluation and acceptance decisions due, December 31, 2019

Step 3 – Data release, the start of the challenge, January 13, 2020

Step 4 – Deadline for participants to submit final solutions and 
reports, May 3, 2020

Step 5 – Publication of final results, May 15, 2020

Data Mining Challenge Proposal Format

Proposals should answer the following questions:

- Who are the organizers? Provide the names, affiliations, contact 
information, and brief bio of the organizers.

- Which sector/application domain is considered in the proposed Data 
Mining Challenge?

- What kind of data set will be provided? Please describe its essential 
characteristics. If there are privacy concerns regarding the data, 
please explain in detail. The KnowledgePit Platform may provide 
additional services regarding the anonymization of the data. For more 
information, please contact the platform staff (contact at knowledgepit.ml).

- What are the considered task and the proposed evaluation metrics?

- What prizes will be offered to winners?

Deadline for submissions of proposals: December 20, 2019. Please send 
the proposals to: contact at knowledgepit.ml

FedCSIS 2020 Data Mining Challenge Co-Chairs

Piotr Biczyk, QED Software (piotr.biczyk at qed.pl)

Andrzej Janusz, Warsaw University & QED Software (andrzej.janusz at qed.pl)

More information about the hpc-announce mailing list