A seasonal ARIMA model based on the gravitational search algorithm (GSA) for runoff prediction

Abstract The prediction of river runoff is crucial for flood forecasting, agricultural irrigation and hydroelectric power generation. A coupled model based on the Gravitational Search Algorithm (GSA) Seasonal Autoregressive Integrated Moving Average (SARIMA) proposed to address non-linear seasonal features data. GSA has a significant local optimisation capability, while SARIMA allows real-time adjustment using historical data suitable analysing time series with variations. Consequently, GSA-SARIMA was developed applied Xianyang section Wei River. results suggest that achieves linear correlation coefficient 0.9351, Nash efficiency 0.91, mean relative error 6.57 root square 0.21. All evaluation indicators this outperform other models developed, its application actual feasible, which creates new path prediction. HIGHLIGHTS Mann-Kendall trend test ascertain separation point between training datasets. It avoids too little in set, effectively improving generalisation model. an improvement ARIMA convenient algorithm applicable parameter search optimization great global capability.