High-Precision Debris Flow Scale Prediction Model Based on CIHHO Algorithm Combined With Multilayer Perceptron Neural Network

Abstract

This study focuses on improving the precision and global search capabilities of the Harris Hawk Optimization algorithm when dealing with debris flow disasters. The enhanced algorithm, called Cubic map Improved Harris Hawk Optimization (CIHHO), incorporates a BP neural network and integrates Grey relational analysis and Spearman correlation analysis to address the factors influencing debris flow disasters effectively. Additionally, The CIHHO-BP model was tested on 92 debris flows in Beichuan County and was found to be more accurate than a standalone BP neural network model in predicting the size of the debris flow scale, and the CIHHO-BP model outperformed a model that combined the Harris Hawk Algorithm (HHO), Genetic Algorithm (GA), Coyote Optimization Algorithm (COA), and Particle Swarm Algorithm (PSO) with the BP neural network. The research demonstrates that the enhanced Harris Hawk optimization algorithm, which integrates a cubic map to optimize the initial population and updates the iterative formula E, effectively resolves the issue of local optimization and inadequate global search capability in the Harris Hawk algorithm. The R2 of the debris flow scale prediction model optimized by the enhanced Harris Hawk Optimization algorithm combined with the BP neural network has reached 0.974, demonstrating enhanced stability in the prediction of the scale of small and medium-sized debris flows. The combined prediction model offers a novel approach to the prediction of debris flow scale, providing a valuable reference for the prevention and control of debris flows and machine learning algorithms.

Keywords

• Debris flow scale
• chaotic mapping
• BPNN
• improved Harris Hawk optimization algorithm
• correlation analysis