Published in Expert Systems with Applications — Heterogeneous transfer learning for infectious disease forecasting (2026)

A heterogeneous transfer learning framework that pre-trains on data from dissimilar countries to address the single-country data limitation in epidemic forecasting, validated on COVID-19 data from 30 countries. Sole author: Donghyun Lee.

• 1,143 days of COVID-19 data across 30 countries · Temporal Convolutional Network (TCN) · source countries selected by Dynamic Time Warping (DTW) distance
• 40% RMSE reduction over the individual TCN baseline, with an R² of 0.82
• Splitting sources into DTW-distance quartiles, the most dissimilar group reduced RMSE by 12.1% versus the most similar (p = .008)

DOI

R&D Agreement Signed with Green Vet Inc.

AI Korea Inc. has signed a mutual R&D agreement with Green Vet Inc., a pet healthcare subsidiary of Green Cross, in veterinary AI diagnostics.

• Joint AI research in companion-animal diagnostic testing — including veterinary microbial diagnostics and antibiotic prescription
• Joint development and validation of AI algorithms
• Sharing of research outcomes and discussion of commercialization paths
• Utilization of joint R&D results and ongoing improvement of commercialized technology

Published in J. Cleaner Production — Reinforced explainable AI for algal bloom forecasting (2025)

A reinforced explainable AI for algal-bloom forecasting under climate change, using a multi-run CAM (class activation mapping) approach. Corresponding author: Donghyun Lee.

DOI

Published in J. Cleaner Production — Reliable AI for PM uncertainty quantification (2024)

A reliable AI framework that quantifies uncertainty in deep-learning particulate-matter predictions, providing evidence for policy and operations. Corresponding author: Donghyun Lee.

DOI