Flood discharge prediction using improved ANFIS model combined with hybrid particle swarm optimisation and slime...

Due to the disastrous socio-economic impacts of flood hazards and estimated rise of its occurrences in the near future, there has been an increase in the importance of flood prediction worldwide. Artificial intelligence (AI) models have contributed significantly by giving cost-effective solutions for simulating physical processes of flood events and improving accuracy in prediction over the last few decades. This paper presents a novel conjoint model to forecast river flood discharge (Q_FD) considering data from four gauging stations of River Brahmani, Odisha India. The developed hybridised metaheuristic algorithm, i.e. ANFIS-PSOSMA, improves exploration capability of Slime mould algorithm (SMA) by integrating it with particle swarm optimisation (PSO). Performance of novel hybrid model is assessed by utilising quantitative statistical measures like the coefficient of correlation (R²), Nash-Sutcliffe Model Efficiency (N_SE), root mean square error (RMSE), and mean absolute error (MAE). The proposed hybrid ANFIS model using optimisation algorithm provided the best performance values with N_SE of 0.9952, R² of 0.9946, RMSE of 0.0485, and MAE of 0.0265 during training and N_SE of 0.9736, R² of 0.9731, RMSE of 8.4236, and MAE of 4.3197 during testing at Jenapur gauging station, indicating the prospective of utilising the developed models in forecasting flood discharge. The present study's importance lies in integrating several input parameters, and AI algorithms have been utilised for developing flood prediction model. In addition, the attained results indicated that combining the optimisation algorithms with ANFIS enhanced its performance in modelling monthly flood discharge time series.