Team of St Petersburg University students wins third place in the global industrial competition on artificial intelligence

The competition Pump it Up: Data Mining the Water Table, in which students from St Petersburg University are taking part, was proposed by an open platform Taarifa that aggregates data from the Tanzanian Ministry of Water Resources and Irrigation. The developers were asked to predict the performance of water pumps across the country based on extremely small data sets: on average one record per water pump. The urgency of this task stems from the inability to continuously monitor all water pumps in the country due to a lack of infrastructure or staff, or sometimes both. However, a clear understanding of which water points will fail and when can improve maintenance operations and ensure the availability of clean drinking water for communities across Tanzania.

Click here to see the AI developer rankings.

The team "Butsa i Studenti in SpBU" included bachelor’s students of the University Timofei Malov, Dmitrii Vasilev, Artem Peshkov, and Mariia Barkovskaia. Together with Professor Ovanes Petrosian, Director of the Centre for Artificial Intelligence and Data Science at St Petersburg University, they offered their solution to the problem, with which they are currently leading in the ranking of developers from universities, research organisations, and IT companies throughout the world.

The idea behind the University’s solution is based on two principles. The way a water pump is used and the environment in which it operates have a direct impact on its operating time. In the absence of data, the modes of operation can be calculated based on population density, while the environment is determined by weather conditions and related indicators. The constructed approach allows predictions to be made about the current condition of a water pump based on indirect indications of its location, date of installation and its type without any additional information. According to Ovanes Petrosian, Director of the Centre for Artificial Intelligence and Data Science at St Petersburg University and the team’s supervisor, similar principles can be used for other similar predictive analytics tasks in conditions of very small data volume for predictive analytics objects geographically distributed across the country (e.g., pipelines and other engineering equipment).

The platform DrivenData holds online data science competitions in which anyone can test their analytical skills by solving real industrial cases. Once the data is published by the competition organisers, any team can use it to build a model, train AI algorithms and submit their solution to the world ranking. Contestants are provided with a training dataset and a set of test data is also published, for which contestants are supposed to make predictions using machine learning methods. Contestants send their predictions via an automated system to the organisers, who already compare them with the actual values. At the end of the competition, the team with the best results is declared the winner.

Ovanes Petrosian, Professor of St Petersburg University, noted that the complexity of the task the students were solving was to find a fundamentally new scientific approach to data analysis and a new algorithmic solution to the problem. ‘Students from St Petersburg University are in the top 3, on a par with a developer from Electronic Arts and a professor from Queen’s University at Kingston, which testifies to the high quality of the University’s students. In order to cope with tasks of this level, it is necessary to have excellent knowledge of data science, statistics and programming, and the ability to use modern new tools for data processing. No less important is the study of modern literature on the topic of the research,’ said Ovanes Petrosian. 

St Petersburg University and the Centre for Artificial Intelligence and Data Science at the University support students participating in industrial competitions in the field of artificial intelligence. According to Order No 1954/1 dated 27 February 2023, an annual scholarship competition is held for them and the University invites talented students to apply for the scholarship. 

The Pump it Up: Data Mining the Water Table competition will run until October 2024, with new and existing participants able to upload new solutions to the problem. The members of the Butsa i Studenti in SbBU team plan to propose more efficient solutions and compete for first place in the ranking. The team that wins gold will be included in the world ranking of leading experts in artificial intelligence. ‘The creation of a top model is an insight, a new statistical approach or a tool for analysing new data that will make it possible to fulfil the mission of solving some of the world’s biggest social challenges in a more efficient and sustainable way,’ emphasised Ovanes Petrosian, Professor of St Petersburg University.