Data Fusion Contest 2022 is a machine learning competition with a unique dataset and tasks: to find the same users by their behavior on transaction data and clickstream.
(sending applications)
February 3 - April 3, 2022
Selection of final solutions
April 4, 2022, until 12:00
Verification of solutions,
determination of finalists
April 4 - April 14, 2022
Solemn awarding of the contest finalists
April 15, 2022
Total award fund ₽2 000 000
Award fund for Task № 1 Matching
1 450 000 ₽
Award fund for Problem № 2 Puzzle
300 000 ₽
Award fund of special nominations
250 000 ₽
Main tasks
Task No. 1
НBuild an algorithm to solve a matching problem. According to transactions and clickstream data, build impersonal vectors for users. Then, based on the created vectors, match each vector of the Bank's client's transactions with a vector based
on click streaming from a partner and rank them according
to the probability of matching.
Metric R1
Solution format Container with the code
1 450 000 ₽
Task No. 2
Build an algorithm that solves the task of matching in the case when all the candidates for the pairs are known in advance, but the pairs themselves are not provided.
Participants are to solve the same correlation task,
but in a form close to doing a puzzle.
Metric R1
Solution format Tagged file
300 000 ₽
Additional tasks
and activities
Nomination No. 2
The award for the most interesting, breakthrough and non-standard solutions to the contest tasks. Research novelty, engineering efficiency, and uniqueness of the approach are assessed. The winners of the nomination are determined by the jury.
100 000 ₽
Educational task
An exercise for those who want to learn how to work with industrial transaction data: create an algorithm capable of predicting the client’s possession of higher education. Gifts
Regular activity
Regular meetups, guest lectures, workshops, and tool reviews for the participants of the contest. Data Fusion 2022 is also a great opportunity to improve your data analysis skills.
Nomination 1

The award for the best public solutions opened before the end of the contest. The winners of the nomination are determined by the jury, but the voting will be based on the reactions and support of the materials by the participants themselves on the contest platform.

150 000 ₽
The uniqueness of the dataset
With the help of Platforma tools, and data exchange secure matching technology, a synthetic dataset was formed on depersonalized data from VTB and Rostelecom.
At the same time, the data was generated in such a way as to preserve the necessary information about user behavior in order to solve the problem.
Data Fusion 2021


How will the solutions of the contest be assessed?
The solutions of the contest tasks (1-3) are assessed automatically, based on comparison with the true test marks, which are available only to the organizers. Applications for special nominations are evaluated by the jury consisting of the contest organizer’s representatives, invited experts and researchers after the decision-making process is complete.
When will the winners of the contest be determined?
The winners of the contest will be determined between April 3 to April 14, 2022.
How are the final solutions chosen?
In each task, you can choose up to 2 of your final solutions. Throughout the contest, you can choose the final solutions, as well as change your choice. If the final solutions have not been selected, then the 2 best solutions are automatically selected on the public part of the test data. The results on public test data are available during the course of the contest, but are not used to determine the winners of the contest.
When does the registration for the contest end?
You can register for the contest at any time until April 3, 2022, 23:59:59 Moscow time.
Will there be an award ceremony for the winners?
Yes. The award ceremony will take place within the framework of the Data Fusion Conference on April 15, in Skolkovo. During the ceremony, the winners of the contest will be announced and the awards will be presented. The finalists will be invited to the award ceremony in person.
Is it possible to solve several tasks at once?
Yes. It is allowed to take part in one task or activity, as well as in any possible combination of them.
Is it possible to participate in teams?
Participation in all tasks is possible in teams of 1 to 4 people.
How are the winners of the contest tasks determined?
In each of the tasks that a participant solves, it is necessary to choose up to 2 of his final solutions. These solutions will be launched and assessed on closed private test data. It is the result on private data that is considered the official final result of the contest. The winners in each of the tasks are the participants or their teams who have received the best results in the rating of the final results.
How to participate in the Data Fusion Contest 2022?
To participate, you must register on the contest platform. Then you need to develop your solution and upload it to the contest platform in accordance with the rules. Solutions can be improved and uploaded throughout the contest.