Dongwoo Kim, a PhD student, and Jaeho Lee, a master’s student in POSTECH Electrical Engineering (advisor: Professor PooGyeon Park), won the Grand Prize and KRW 15 million in the Steel Raw Material Price Trend Prediction category of the POSCO Industrial-AI Solution Challenge.

Their team, Serio, applied a dual-stage attention-based recurrent neural network (DA-RNN), automatically learning the importance of each input and the length of historical data used for training to predict future iron ore prices efficiently.

After taking first place in online preliminaries based on iron ore prices released weekly from mid-July through late August, the team won the Grand Prize following additional validation and a presentation.

The challenge was the steel industry’s first AI solution competition. It sought creative solutions for three tasks: predicting steel raw material price trends to reduce purchasing risks from global market uncertainty; predicting swell-wave occurrence at ship unloading docks to prevent unloading disruption; and predicting cafeteria meal attendance to minimize food waste.

A total of 451 participants from 206 teams competed through June preliminaries and September finals, with the top three teams per task, nine in total, advancing.