Outstanding Paper Award 2025 in the Database Field, Journal of KIISE

At the Korea Computer Congress 2025 (KCC 2025), held over three days from 2 to 4 July, the paper 'A GRU-based Time Series Forecasting Method Using Patching' by master's graduate Kim Yun-young (supervisor: Jung Sung-won), published in the database field of the Journal of KIISE in 2024, was selected as an outstanding paper among all papers published in that field in 2024 and received the Outstanding Paper Award.

The paper is the result of research written by Kim Yun-young (master's graduate) under the supervision of Professor Jung Sung-won.

▲ From left: Kim Yun-young (master's graduate), Jung Sung-won (supervisor)

A variety of neural network based models — RNN, MLP, CNN, transformer and others — have been studied for time series forecasting, but the transformer family is limited in practical application by its complexity and heavy computational resource consumption, while simple MLP models struggle to learn sufficient information because of their structural constraints. To overcome these limitations, this research proposes PatchTSG, a Gated Recurrent Unit (GRU) based time series forecasting model with patching applied. PatchTSG shortens the input length through patching and combines it with a GRU to secure efficiency, and showed better performance than a single linear model on many benchmark datasets. By shortening training time by up to 82% and inference time by up to 46% compared with transformer-based models, it was demonstrated to be a lightweight time series forecasting model applicable to a variety of real settings such as traffic flow and energy demand forecasting.