Evaluation of Ensemble Learning Methods for Flood Prediction Taskin Tahmini I in Topluluk grenme Y ntemlerinin Degerlendirilmesi

Abdulsahib Al-Qazzaz-L Z. A., PAŞA Y., Pashaei E.

33rd IEEE Conference on Signal Processing and Communications Applications, SIU 2025, İstanbul, Türkiye, 25 - 28 Haziran 2025, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu66497.2025.11112464
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: Ensemble learning, Flood prediction, Machine learning, Majority Voting
  • İstanbul Gelişim Üniversitesi Adresli: Evet

Özet

Flooding in India, exacerbated by heavy rainfall from tornadoes, has caused significant damage in recent years. This study aims to enhance flood prediction using machine learning techniques. It evaluates the performance of ensemble learning methods - Voting, Blending, Stacking, Bagging, and Boosting - alongside traditional models such as k-nearest neighbors (KNN), decision tree (DT), random forest (RF), support vector machine (SVM), logistic regression (LR), and a Gated Recurrent Unit (GRU) based deep learning model. Notably, this is the first study to comprehensively compare various ensemble methods for flood prediction. Models were assessed based on accuracy, recall, precision, F1-measure, and ROC score. The findings demonstrate that ensemble methods, particularly Stacking, outperform individual models. While the KNN model showed the lowest accuracy at 78%, the GRU model achieved 89%, and the Stacking ensemble delivered the highest accuracy at 92%. These results underscore the potential of ensemble learning to improve flood prediction, offering valuable insights for disaster preparedness and mitigation strategies