27 August 2019 – Computer Science PhD students Chen Yizhou, Yu Haibin and Dai Zhongxiang won the world championship title at the global finals of the JDD-2018 Global Digitalization Challenge held from 12 to 13 January this year in Beijing, China.
The team was one of 1580 teams to compete in the population census data competition organised by JD.com – a Chinese e-commerce company. The winning team won a cash prize of ¥500,000 (S$100,000) and each member was offered full-time job opportunity with JD.com.
In this competition, teams were challenged to create an AI model that could accurately predict changes in population data of different Chinese cities, districts and counties. Teams were given data on mobile device users of different districts and cities to train their AI models.
The competition is divided into four main divisions – the mainland China division, Hong Kong division, USA division and Israeli division. In the tryouts and division finals, teams are ranked by their performance in the division competitions. The best two performing teams of each division competed in a 24-hour offline challenge to perfect their model and develop a business plan for their model.
According to team member Zhongxiang, the team was able to identify and remove irregularities in the given dataset using an algorithm they developed. This allowed the team to remove a significant number of prediction errors that their model was generating. The team’s choice to use a mixture of different models also helped the team create a system that was able to generate impressive predictive results.
“At the initial stage of the competition, our results were not good and it seemed as though we may not proceed to the finals. We chose not to give up and chose to explore other methods such as trying out other types of models, and to study the data given,” said Zhongxiang. “Thankfully, our efforts were not in vain and we ended up leading on the leaderboard, as well as winning first place.”
Yizhou and Zhongxiang are currently in the second year of their PhD studies while Haibin is in his fifth year. All three members are under the supervision of Assistant Professor Bryan Low and their research area is in machine learning.
Ji Qi Zhi Xin, 16 January 2019