Flood-prone area mapping using a synergistic approach with swarm intelligence and gradient boosting algorithms.

Area of Science:

• Environmental Science
• Geographic Information Science
• Artificial Intelligence

Background:

• Accurate flood susceptibility mapping (FSM) is crucial for effective flood risk management.
• Existing FSM methods often lack optimal hyperparameter tuning for machine learning models, leading to reduced accuracy.
• There is a need for advanced techniques to enhance FSM precision for better decision-making.

Purpose of the Study:

• To introduce and evaluate a novel approach for improving FSM accuracy.
• To optimize the CatBoost machine learning algorithm using swarm-based metaheuristic algorithms (Zebra Optimization Algorithm - ZOA, Whale Optimization Algorithm - WOA).
• To enhance FSM in Shushtar County, Iran, by applying optimized CatBoost models.

Main Methods:

• Utilized 13 flood-influencing parameters and flood occurrence points as input data for FSM.
• Applied the CatBoost algorithm and optimized versions using ZOA (CatBoost-ZOA) and WOA (CatBoost-WOA).
• Evaluated model performance based on accuracy metrics for flood susceptibility maps.

Main Results:

• The standard CatBoost model achieved 84.2% accuracy.
• The CatBoost-WOA model reached 85% accuracy.
• The CatBoost-ZOA model demonstrated the highest accuracy at 87.2%, showing a 3.0% absolute improvement over the non-optimized CatBoost model.

Conclusions:

• Integrating swarm-based optimization algorithms with machine learning significantly enhances FSM accuracy.
• The optimized CatBoost models (CatBoost-ZOA and CatBoost-WOA) offer a more accurate and reliable approach to FSM.
• This non-structural approach provides valuable insights for flood management and decision-making.