Present study makes an effort to estimate the runoff as a function of rainfall, temperature, and humidity of the Barak River. Two approaches of artificial neural network (ANN) (i) feedforward backpropagation neural network (FFBPNN) and (ii) layer recurrent neural network (LRNN) are used to predict runoff during the monsoon period. The evaluation criteria considered for both of the models are the mean square error (MSE), root mean square error (RMSE), and coefficient of determination (R2). FFBPNN performs best with architecture 4-6-1 following Tan-sig transfer function, which possesses MSE training and testing values 0.00012 and 0.00116, RMSE training and testing values 0.01072 and 0.03401, and R2 training and testing values 0.9714 and 0.9645. LRNN performs best with architecture 4-7-1 following Tan-sig transfer function, which possesses MSE training and testing values 0.00021 and 0.00069, RMSE training and testing values 0.01449 and 0.02617, and R2 training and testing values 0.9467 and 0.9380. Further, the comparison of FFBPNN, LRNN, and multiple linear regression techniques (MLR) is done and the result shows that all the approaches perform best and gives good R2 values. Findings of this study could be used for planning and management of hydraulic structures in neighborhood of watershed.
CITATION STYLE
Annayat, W., Gupta, A., Prakash, K. R., & Sil, B. S. (2021). Application of Artificial Neural Networks and Multiple Linear Regression for Rainfall–Runoff Modeling. In Lecture Notes in Networks and Systems (Vol. 134, pp. 719–727). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-15-5397-4_73
Mendeley helps you to discover research relevant for your work.