Utilizing Strategic Pre-training to Reduce Overfitting: Baguan – A Pre-trained Weather Forecasting Model
Published in KDD 2025 (31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining), 2025
This paper introduces Baguan, a novel data-driven model for medium-range weather forecasting, built on a Siamese Autoencoder pre-trained in a self-supervised manner. The model effectively mitigates overfitting through strategic pre-training and demonstrates superior performance in global forecasting and downstream tasks.
Recommended citation: Peisong Niu, Ziqing Ma, Tian Zhou, Weiqi Chen, Lefei Shen, Rong Jin, Liang Sun. (2025). "Utilizing Strategic Pre-training to Reduce Overfitting: Baguan -- A Pre-trained Weather Forecasting Model." Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2025). https://arxiv.org/abs/2505.13873