Comparison of the Performance of PSO and GA Algorithms in Predictive Modeling of Flood-Related Deaths in Boma

This study examines river dynamics and flooding in the town of Boma, Democratic Republic of Congo, where vulnerability to flooding is increased by climate change and anthropogenic pressures. This study aims to address gaps in flood-related fatality prediction by developing a predictive model incorporating the interaction between the Congo River water level and the Kalamu River discharge. The objectives include the use of a generalized linear model (GLM) with a Poisson distribution, combined with optimization algorithms such as particle swarm optimization (PSO) and genetic algorithms (GA). The methodology relies on the collection of historical data on water levels, discharges, rainfall, and fatalities, followed by rigorous data analysis using preprocessing and optimization techniques. The results show that PSO outperforms GA in terms of convergence speed and efficiency, achieving a better fitness value. Fitness values reveal an RMSE of ۸.۳۷, an MAE of ۶.۴۲, and an R² of -۴.۰۴, indicating significant inaccuracies in the forecasts. Simulations reveal a direct relationship between water level, discharge, and deaths, highlighting the importance of these interactions for risk management. These results provide valuable tools for infrastructure planning and raising awareness of the impact of floods on vulnerable populations, thus contributing to more effective prevention strategies.

کلیدواژه ها:

Floods ، Predictive modeling ، River dynamics ، PSO and GA optimization algorithms ، Boma