AUTOMATIC FEATURE SELECTION FOR STORM SURGE FORECASTING

  • Thi Hien Nguyen Institute of Information and Communication, Le Quy Don Technical University
  • Thi-Thu-Hong Phan Department of Artificial Intelligence, FPT University, Da Nang
  • Sooyoul Kim Centre for Water Cycle, Marine Environment and Disaster Management, Kumamoto University
Keywords: Feature selection, genetic algorithm, wrapper method, storm surge

Abstract

Accurate prediction of storm surge events is crucial due to their severe nature, with the potential for widespread flooding and destruction. Timely and precise predictions are necessary to facilitate effective evacuation orders and enable prompt emergency responses.
Machine learning has consistently demonstrated its superiority over traditional models in storm surge forecasting in terms of both accuracy and timeliness. However, the accuracy of machine learning models in storm surge prediction greatly relies on selecting the appropriate
input features for training. To address this issue, this paper proposes a feature selection method that utilizes a GA to determine the optimal set of input features for machine learning models in storm surge prediction. Through experiments conducted on storm surge data from the Tottori coast of Japan and Vietnam, the proposed approach exhibits significant improvements in the accuracy of machine learning models for storm surge prediction.

điểm /   đánh giá
Published
2025-08-28
Issue
Vol. 14 No. 01 (2025)
Section
Bài viết