Comparison of Artificial Neural Network and Support Vector Machine for Long-Term Runoff Simulation

Abstract

Simulation of runoff from a river catchment is a very difficult task and it involves a lot of data which need to be considered. However, the modelling is very essential to forecast the patterns of future runoff by observing and analysing previous patterns of runoff, based on the rainfall. This study presents the evaluation of rainfall-runoff modelling for the long-term runoff series using Artificial Neural Network (ANN) and Support Vector Machine (SVM). Both models are trained and validated using the data series of current and nine (9) antecedent rainfall. During the validation, the SVM model is better in the performance as compared the ANN model, with the R and RMSE values are 0.529-0.711 and 14.27-52.55 mm, respectively. However, the SVM model is underestimated for the peak discharge. Both models have the ability to derive the relationship between the inputs and outputs of the rainfall-runoff process.

Comparative Analysis of Rainfall-Runoff Modeling Using Support Vector Machines for Two Dams in Uttarakhand

“…The algorithms that are capable of learning inductively by using the examples have been used to various difficult, nonlinear, real-world problems of practical interest [8]. During the validation, the SVM model is better in performance as compared to other models [9]. Recently many researchers applied SVM successfully in rainfall-runoff modeling [10]- [12].…”