Prediction of seasonal maximum wave height for unevenly spaced time series by Black Widow Optimization algorithm

The present study aimed to predict the maximum seasonal wave height by new integrative data driven methods. For this purpose, two data-driven techniques, that are, the Adaptive Neuro-Fuzzy Inference System (ANFIS) and the Support Vector Regression (SVR), were applied, and a BWO algorithm was used as an integrated method (ANFIS-BWO and SVR-BWO). In addition, the Particle Swarm Optimization (PSO) algorithm was used as a method integrated with SVR and ANFIS (SVR-PSO and ANFIS-PSO) to compare the performance of the newly developed methods (ANFIS-BWO and SVR-BWO). The wave data were collected in different seasons by a buoy station deployed in the southern Baltic Sea by the Institute of Hydro-Engineering of the Polish Academy of Sciences. Seasonal simulations were performed to investigate the effect of seasons on the maximum wave height. The wave data constituted an unevenly spaced time series. The maximum wave height was modeled using the maximum wave height period (T_max), the significant wave height (H_s), the significant wave period (T_s), and time steps (Δt). The results showed that the application of BWO and PSO algorithms increased the accuracy of ANFIS and SVR by about 18.45%. Moreover, the results show that PSO increased the accuracy of ANFIS and SVR by about 17.98% and 21.59%, respectively. The results of different runs indicated that the BWO is more stable to reach the global solution than PSO. The results also show that show that SVR-BWO is the most accurate model.