Assessing machine learning models for streamflow estimation: a case study in Oued Sebaou watershed (Northern Algeria)

Artificial intelligent systems optimized by metaheuristic algorithms and teleconnection indices for rainfall modeling: The case of a humid region in the mediterranean basin

“…Additionally, 100% of the data from training and validation were used in the testing stage. It should be noted that the choice of dividing the data into 70/30% is the most commonly used approach in recent applications [ 64 , [77] , [78] , [79] , [80] , [81] ]. Different algorithms, such as Levenberg-Marquardt, Quasi-Newton Resilient Backpropagation, Scaled Conjugate Gradient, and Variable Learning Rate Backpropagation, can be used to train MLP networks.…”

“…The LM algorithm makes it possible to obtain a numerical solution to the problem of minimizing a function, often nonlinear and dependent on several variables. It is more stable than Gauss-Newton, and it may find a reasonable solution even when the initial solution starts far from the optimum [ 79 ].…”

Using Deep Learning Algorithms for Intermittent Streamflow Prediction in the Headwaters of the Colorado River, Texas

“…ELM reduces the computational time and enhances the generalization ability of the single-layer Artificial Neural Network (ANN) model [58,60,61]. ELMs have gained popularity in the hydrologic literature [62][63][64][65][66][67] and are established as fast and effective streamflow forecasting models [37,[68][69][70][71][72][73].…”

“…ELM reduces the computational time and enhances the generalization ability of the single-layer Artificial Neural Network (ANN) model [58,60,61]. ELMs have gained popularity in the hydrologic literature [62][63][64][65][66][67] and are established as fast and effective streamflow forecasting models [37,[68][69][70][71][72][73]. An ANN architecture consisting of one input layer and L hidden neurons with an activation function called g(x) and a bias term (B), is presented mathematically as Equation ( 1):…”

Near-real-time satellite precipitation data ingestion into peak runoff forecasting models

“…It is common to find new studies using Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), and Random Forest (RF) (Bhattacharya and Solomatine, 2005;Dibike et al, 2001;Hosseini and Mahjouri, 2016;Mosavi et al, 2018;Muñoz et al, 2021Muñoz et al, , 2018Solomatine and Dulal, 2003;Tongal;Booij, 2018;Young et al, 2017). Comparisons between these techniques have demonstrated that RF produces promising results in comparison to SVMs and ANNs, especially because RF is capable to deal with small size samples and complex data structures (Abda et al, 2022;Galelli;Castelletti, 2013;B. Li et al, 2016;Papacharalampous and Tyralis, 2018;Solomatine and Dulal, 2003).…”