A hybrid DE optimized wavelet kernel SVR-based technique for algal atypical proliferation forecast in La Barca reservoir: A case study

The atypical proliferation of algae is a consequence of eutrophication, a phenomenon responsible for the deterioration of reservoirs and lakes. Its growth over the last few decades forced different administrations to adopt different solutions, including forecasting and management, with the help of mathematical models. This article presents a model of eutrophication of reservoirs based on a new methodology called multiscale Mexican Hat wavelet as the kernel function for the support vector regression (SVR) method and differential evolution (DE) optimization technique to estimate the abnormal proliferation of algae from physicochemical and biological variables. The present method implies the optimization of the SVR hyperparameters during the training process. In addition, five other SVR models with different nuclei (linear, quadratic, cubic, sigmoid and radial base function) and random forests (RF) were adjusted to experimental data for purposes of comparison. In addition to successfully predicting atypical algae growth (determination coefficients equal to 0.88 and 0.93), the model shown here can establish the importance of each biological and physicochemical parameter of improved algae growth. Finally, the main conclusions of this research work are presented. •A hybrid DE/SVR-based model is a good predictive model of the phosphorus and chlorophyll presences.•Algal atypical proliferation is dangerous for environment in reservoirs.•The results of this new model for the eutrophication evaluation is compared with the RF model tree.•The obtained correlation coefficients of hybrid DE/SVR-based models are about 90%.•The biological and physical–chemical input variables in this study are studied in depth.