Hybrid PSO and GA for Neural Network Evolutionary in Monthly Rainfall Forecasting

Accurate and timely weather forecasting is a major challenge for the scientific community in hydrological research such as river training works and design of flood warning systems. Neural Network (NN) is a popular regression method in rainfall predictive modeling. This paper investigates the effectiveness of the hybrid Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) evolved neural network for rainfall forecasting and its application to predict the monthly rainfall in a catchment located in a subtropical monsoon climate in Guilin of China. Our methodology adopts a hybrid Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) for the automatic design of NN by evolving to the optimal network configuration(s) within an architecture space, namely PSOGA–NN. The PSO is carried out as a main frame of this hybrid algorithm while GA is used as a local search strategy to help PSO jump out of local optima and avoid sinking into the local optimal solution early. The proposed technique is applied over rainfall forcasting to test its generalization capability as well as to make comparative evaluations with the several competing techniques, such as GA–NN, PSO–NN and NN. The experimental results show that the GAPSO–NN evolves to optimum or near–optimum networks in general and has a superior generalization capability with the lowest prediction error values in rainfall forecasting. Experimental results reveal that the predictions using the GAPSO–NN approach can significantly improve the rainfall forecasting accuracy.