Blog

We’ve got a winner! ibi sponsors data science competition By Media Shower

  |   RSS Feeds

ibi recently organized a machine learning competition in association with the IEEE BigData 2020 Conference and QED Software, and we’ve got big news about the winner. The challenge The best challenges are the ones that have real-world implications, so ibi designed the competition around a real-world customer support challenge. Competing teams had to create a
The post We’ve got a winner! ibi sponsors data science competition appeared first on ibi.

ibi recently organized a machine learning competition in association with the IEEE BigData 2020 Conference and QED Software, and we’ve got big news about the winner.

The challenge

The best challenges are the ones that have real-world implications, so ibi designed the competition around a real-world customer support challenge. Competing teams had to create a machine learning model to accurately predict which customer support tickets would get escalated by the customer. For instance, a customer with a critical deadline may require dedicated support for a period of time. When this occurs, an escalation arises. Being able to anticipate when this will happen helps ibi to provide better customer support and a better experience overall.

Here at ibi, we wanted to leverage machine learning and AI as a mechanism to continually improve our ability to predict and manage customer support escalations. Every service escalation is disruptive, so the goal of the contest was to create a model to proactively spot an escalation by determining the cause, whether it’s a production outage, duration, technical proficiency, or other issue.

This data science competition, which will be featured at the IEEE Big Data 2020 conference in December, allowed teams to create data science models using machine learning along with anonymized data to solve the challenge.

The participants

The 2020 IEEE BigData Cup event used KnowledgePit, an online platform that connects companies with data scientists and facilitates all the necessary operations that are required to prepare the problem and data set, run the competition, tally the results, and choose the winner(s). It also helps the problem owner to adopt the winning solution(s) in their business.

The IEEE BigData 2020 Cup attracted 250 teams from 50 countries to develop new machine learning models in an effort to predict escalations in customer support. For participants, this was an exciting competition and a chance to prove themselves.

Here’s how it worked:

  • The registered teams were supplied with a dataset comprising about 10 percent of the full dataset their model would ultimately be tested against during the competition
  • ibi supplied the baseline data model, created by the ibi team to give competitors a point of reference on how a model that was created quickly performs. The team could see the code for the baseline model
  • The teams created code on their own, and then tested code performance on the anonymized data ibi provided. How did it perform compared to the baseline?
  • Then the teams submitted their models for final testing and final scoring

Competition was fierce, and ibi was excited to see so many promising and talented people working on the challenge.

We’re pleased to announce that the winner of the challenge was Peter Klimov and team. Congratulations!

Comments from the winners

Beyond the prestige and cash prize for winning the competition, the winning team noted some other benefits as well. When asked about interesting or surprising things they had learned by participating, they replied:

“In the process of solving this problem, we studied and used a number of interesting architectural approaches related to the general construction of our solution. We were also interested in gaining experience in this kind of task, combining different types of features with local time series and an interesting context for solving a problem related to escalation.

“Before this competition, we had not thought that such problems could have interesting interpretations in the context of machine learning. We are very pleased to have taken part in this competition and would be happy to take part again in contests of this format.”

Everybody wins

ibi feels strongly that competitions such as this offer a win-win scenario. The two participating, winning parties are the business community and community of researchers and practitioners working with data-related problems (data scientists). For participants, it is an exciting chance to showcase their skills, and for the company that provides the problem and data, it is a novel and potentially beneficial approach to solve an important business problem.

Sponsoring the competition is part of ibi’s continuing support of data science and machine learning. Data is in our DNA, and we believe that the future of data science is bright with the continuing evolution of artificial intelligence and machine learning capabilities. Forward-thinking organizations are already embracing future-facing technologies that enable intelligent decision-making and informed problem-solving.

The post We’ve got a winner! ibi sponsors data science competition appeared first on ibi.