Evaluation of Decision Tree and Support Vector Machine Classifiers in Comparison for Flood Prediction

Abstract

Protection from floods is one of the most significant activities aimed at risk reduction. Flood prediction and disaster preparedness have relevance for reducing the association between floods and people and infrastructure. The goal of this study was to find out how well the Support Vector Machine (SVM) and Decision Tree classifiers work at predicting flooding based on different environmental and infrastructure factors. Data collection had been done through gathering 140 samples with a total of 21 variables where analysis had been done in order to identify significant contributory factors: topography, urbanization, and climate change. The results show that an SVM model was capable of achieving an accuracy of 91%, while the Decision Tree classifier did much better at an accuracy of 95%. The decision tree model was also more precise in flood prediction (1.00) and recall for non-flood cases (1.00), while for both models, the recall for flood cases was the same (0.88). This indicates that both the models had some false negatives for floods. The current study focuses more on machine learning applications and disaster readiness in flood risk assessment for better and more effective mitigation.