Flood Area Prediction using a Stacked Ensemble of Tree-Based Algorithms
Floods cause significant loss of life, property damage, and long-term socioeconomic disruptions, with over 100 annual deaths globally. This research addresses the drawbacks of the existing models, such as overfitting effects, inadequate dataset and limited study areas through the adoption of a stack...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Sakarya University
2025-06-01
|
| Series: | Sakarya University Journal of Computer and Information Sciences |
| Subjects: | |
| Online Access: | https://dergipark.org.tr/en/download/article-file/4548282 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Floods cause significant loss of life, property damage, and long-term socioeconomic disruptions, with over 100 annual deaths globally. This research addresses the drawbacks of the existing models, such as overfitting effects, inadequate dataset and limited study areas through the adoption of a stacked ensemble-based model. The model contained five different tree - based models namely hoeffding tree, decision tree, functional tree, reduced error pruning (REP) tree and decision stump algorithms. The model was implemented as a system using MATLAB Simulink, version 2020a on laptop with 4GB Memory. Experimental results indicate that REP Tree performed better than other four individual tree algorithms with accuracy of 98.74%, 97.81% and 97.43% for Dataset A, Dataset B and Dataset C respectively. For Dataset A, stacked ensemble model performed better than single algorithms with accuracy, precision, specificity, f1score and recall of 99.62%, 99.51%, 99.51%, 99.63% and 99.73% respectively. For Dataset B, stacked ensemble model also performed better than single algorithms with accuracy, precision, specificity, f1score and recall of 98.45%, 99.11%, 98.12%, 97.37% and 99.06% respectively. For Dataset C, stacked ensemble model performed better than single algorithms with accuracy, precision, specificity, f1score and recall of 98.75%, 99.25%, 99.64%, 99.90% and 99.24% respectively. Our model’s 99.62% accuracy on Dataset A demonstrates potential for integration with real-time sensor networks, enabling scalable flood early-warning systems in vulnerable regions like Lagos and Kuala Lumpur. |
|---|---|
| ISSN: | 2636-8129 |